REDUCING LEAD IN DRINKING WATER:

A BENEFIT ANALYSIS

\ Ronnie Levin Office of Policy, Planning and Evaluation U.S. Environmental Protection Agency EPA-230-09-86-019

Draft Final Report December 1986 (revised Spring 1987)

ACKNOWLEDGEMENTS

Many people contributed to the publication of this document: This analysis would never have been undertaken without Joel Schwartz.

This document would not have been published without the help and encouragement of J'eanne Briskin and Michael B. Cook.

Much appreciation also to J. Michael Davis (ECAO/ORD), Brendan Doyle (OPA/OPPE), Lester Grant (ECAO/ORD), Peter Karalekas (Region I), Peter Lassovszky (ODW/OW), James w. Patterson (Illinois Institute of Technology), Michael Schock (Illinois State Water Survey), and Ethel Stokes (EPA) -- who consistently provided more help than they ever wanted to. Thanks also to Gene Rosov (WaterTest Corporation).

Finally, special thanks to the many people who repeatedly reviewed successive drafts of this report.

TABLE OF O)NTENTS

PAGE

CEAPl'ER I: . INTRODUCTION

I .A. BACKGROUND INFORMATION I-1

I. B. SlMMARY OF REFORT I-2

I. B 1. The Occurrence of Lead in I-5 Public Drinking Water

I.B.2. Benefits of ReducinJ Children's I-1O Exposure to Lead

I.B.3. Blood-Pressure-Related Benefits I-16 and other Adult Health Effects

I.B.4. Benefits of Reduced Materials Darra.ge I-19

I. B. 5. Summary of Annual Benefits of Reduced I-22 Lead in Drinkinq Water

I.C. OOS'IDN CASE S'IUDY I-25

CHAPI'ER II: OCCURRENCE OF IEAD IN DRINKING WATER

II .A. SOURCES OF LEAD IN DRINKIID WATER II-2

II.A.1. Variables Affecting Lead Levels in Drinking Water II-3

II.A.La. Key Water Parameters Affecting the Solubilicy of Lead II-4

II.A. 1.b. Lead Solder II-6

II.A.1.c. 'Lead Pipes II-1O

II .A. 1. d. Other Potential Sources of Lead II-1O

II.A.Le. Other Factors Relating to Lead Contamination Lev~ls II-12

II.A.1.f. Surrarary II-13

II.A. 2. Prevalence of Lead Materials in Distribution Systems II-14 Table of Contents

PAGE

II.B. mTA ON THE OCCURRENCE OF LEAD IN DRINKING WATER II-16

II.B.l. Patterson Study/CUlligan Data on Tap Water II-18

II.B.l.a. Consistency with Other Data II-21

II.B.l.b. Potential Biases in the Data II-25

II.B.l.c. Alternative Analysis of Potential Exposure to Lead in Drinking Water II-32

II.B.2. Lead Contamination in New Hrusing II-38

II.C. ESTIMATED EXPOSURE TO LEAD IN U.S. TAP WATER II-41

II.C.l. Uncertainties and Assurrptions in the Analysis II-42

II.C.2. Calmlations of Exposure to Lead in Drinking Water II-54

II.C.2.a. Estirrate of Exposure to Lead in Drinking Water to Inhabitants of New ~ousing II-55

II.C.2.b. Estirrate of Exposure to Lead in Drinking Water to Inhabitants of Older Hrusing II-57

II.C.2.c. Total Estirrated Exposure to Lead in Drinking Water II-58

CHAPI'ER III: BENEFITS OF REDUCING CHILDREN'S EXPOSURE 'ID LEAD

III.A. PATHOPHYSIOI.OOICAL EFFECTS III-6

III.A. 1. Effects of Lead on Pyrimidine Metabolism III-11

III.A.2. Effects on Heme Synthesis and Related Hematological Processes III-11

III.A.2.a. Mitochondrial Effects III-12

III.A.2.b. Heme Synthesis Effects III-12

III.A. 3. Lead' s Interference with Vitamin D Metabolism and Associated Physiological Processes III-13 Table of Contents

PAGE

III.A.4. Stature Effects III-17

III.A.4.a. Effects of Lead on Fetal GrCMth III-18

III.A.4.b. Effects of Lead on Post-Natal Growth III-.20

III.A.4.c. Sumrra:ry of Stature Effects III-26

III. B. NEUROI'0XIC EFFECTS OF LEAD EXFOSURE III-28

III. B .1. Neurotoxicity at Elevated Blood-Lead Levels III-28

III.B.2. Neurotoxicity at Lo.rler Blood-Lead Levels III-31

III.B.2.a. Cognitive Effects of Lo.rler Blood-Lead Levels III-34

III.B.3. The Magnitude of Lead 's Inpact on IQ III-40

III.C. FETAL EFFEcrs III-42

III.C.1. Assessing the &.nefi,ts of Reduced III-44 Fetal Exposure to Lead

III.D. M:>NETIZED ESTIMATES OF OULDREN'S HEALTH BENEFITS III-49

III.D.1. Reduced Medical Costs III-SO

III.D.2. Costs Associated with Cognitive Damage III-54

III.D.2.a. Corrpensatory Education III-55

III.D.2.b. Effect U[:on Future Earnings III-56

III.D.3. Sumnary of Monetized Benefits III-60

III.E. VALUING HEALTH EFFECI'S: CAVEATS AND LIMITATIONS III-62

III. F. Sl.M1ARY OF ANNUAL MJNETIZED AND N0N--M'.:>NEI'IZED III-67 OIILDREN'S HEALTH BENEFITS Table of Contents

PAGE CHAPl'ER N: HEALTH BENEFITS OF REDUCING LEAD: ADULT ILLNESSES

rv .A. THE REf.ATIONSHIP BEIWEEN BLOOD LEAD LEVELS AND BLO'.)D PRESSURE rv-2

rv.A. 1. Epidemiological Studies of Blood lead Levels IV-3 arrl :Efypertension

IV .A.La. Occupational Studies IV-4

rv.A. 1.b. 01::servational Studies IV-5

IV.A.Le. Pcpulation Studies IV-7

rv.A.2. Mecnanisms Potentially Underlying lead-Induced Hypertension Effects IV-16

IV.A.2.a. Role of Disturbances in Ion TransfX)rt by Plasma Membranes rv-17

IV .A. 2. b. Role of Renin-Angiotensin in Control of Blood Pressure rv-18 and Fluid Balance

rv.A.2.c. Effecc.s of Lead on Vascular Reactivicy IV-22

Iv..A.2.d. Effects of Lead on Cardiac M.lscle rv-23

rv.A.2.e SUrma:ry of Lead-Related Effects on the rv-25 Cardiovascular System

IV .A. 3. Cardiovascular Disease Rates and Water Hardness IV-26

IV.A.3.a Studies of Cardiovascular Disease and Water Hardness IV-27

rv.A.3.b. Lead, Soft Water and cardiovascular Disease IV-30

IV.A.4. Benefits of Reduced Cardiovascular Disease: Reductions in Hypertension and Related Morbidicy and Mortalicy rv-32

IV.A.4.a. Ifypertension rv-35

rv.A.4.b. Myocardial Infarctions, Strokes, and Deaths rv-35

IV.B. LEAD'S EFFECTS UIDN REPRODUCTIVE FUNCI'ION N-40

IV.B.l Est:inatir:g the Population At-Risk for Female Reproductive Effects IV-43 Table of Contents

PAGE IV.C. MJNEI'IZED ESTIMATES OF ADULT HEALTH BENEFITS: REDUCED CARDIOVASOJIAR DISEASE RISK IN MEN IV-45

IV. C. l. Hypertension IV-46

IV.C.2. Myocardial Infarctions IV-47

IV.C.3. Strokes IV-51

IV .C.4. Mortality IV-53

IV. C. 5. Sumrra:ry of Annual Monetized Benefits of Reduced cardiovascular Disease IV-54

IV .D. VALUING HEALTH EFFECTS: CAVEATS AND LIMITATIONS IV-56

IV.E. SUMMARY CF ANNUAL MONEI'IZED AND NONMJNETIZED ADULT HEALTH BENEFITS OF REDUCING EXroSURE TO LEAD IN DRINK[NG WATER IV-6O

CHAPI'ER V: BENEFITS FROM REDUCED MATERIALS mMAGE

V .A. THE CHARACI'ERISTICS OF AGGRESSIVE WATER V-4

V.A.l. Parameters of Water Affecting Corrosivity V-5 ,,~,

V.A.2. The Electroc:nemistry 0£ Corrosivity V-8

V.A.3. Types of Corrosion V-9·

V.A.4. Corrosion Indices V-1O

V.A.5. Plurrbosolvency and Other Factors Detennining Lead Levels in Drinking Water V-12

V.B. DAMAGE TO PUBLIC SYSTEMS FROM INTERNAL O)RRQSION V-15

/ V.B.l. Occurrence of Corrosive Water in the United St~tes V-15

V .B. 2. Corrosion Damage V-2O

V.B.3. Estirrating the Anrual Costs of Corrosive Water V-22

V.B.4. Monetized Benefits of Reduced Corrosion Damage V-29

BIBLIOGRAPHY

APPENDIX A: BOSTON CASE S'IUDY

LIST OF TABLES

PAGE

CHAP1'ER I:

TABLE I-1 SUbstances Included in the 1985 Prcposed I-3 National Prirrary Drinking Water Regulations

TABLE I-2 Estirrates of Annual Per Capita Corrosion I-21 Damage

TABLE I-3 Sumrrary of Estirrated Annual Monetized I-23 Benefits of Reducing Exposure to Lead fran 50 ug/1 to 20 ug/1 (1985 dollars) for Sample Year 1988

TABLE I-4 Surmary of Estirrated Annual Non-Monetized I-24 Benefits of Reducing Exposure to Lead frcm 50 ug/1 to 20 ug/1 for Sample Year 1988

aJAPI'ER I I:

TABLE II-1 Distribution of Culligan Data II-22 {Patterson, 1981)

TABLE II-2 M.micipal Water Samples and Population II-30-31 Percentages fra:n Culligan/Patterson Study (Patterson, 1981)

TABLE II-3 Lead Contamination Levels in Tap Water by II-39 Age of Plumbing (Field Studies)

TABLE II-4 Percentages of Samples Exceeding 20 ug/1 II-44 of Lead at Different pH Levels, by Age of Hcuse Lisr OF TABLES

PAGE

CHAPl'ER III:

TABLE III-1 Blood Lead Levels of Children in III-8 the United States 1976-80

TABLE III-2 Percent of Children Requiring Chelation Therapy III-52

TABLE III-3 Est:inated Anrrual Benefits of Reduced IQ Danage by Using Changes in Expected Future Lifetirre Earnings for Sarcple Year 1988 III-59

TABLE III-4 M:Jnetized Anrrual Benefits of Reducing Children's Exposure to lead Using Alternative Methods for Sarcple Year 1988 III-61

TABLE III-5 SUimary of Annual Monetized and Non-nonetized Children's Health Benefits of Reducing lead in Drinking Water for Sarcple Year 1988. III-68

CHAPl'ER IV:

TABLE IV-1 Benefits of Reducing Strokes IV-52

TABI.E IV-2 Sunmary of Anrrual Monetized Blood­ Pressure Related Benefits of r.o.,,e:i:-ed. MCL for Sarcple Year 1988 IV-;-55

TABLE IV-3 SUima.ry of Annual Monetized and Non­ IV-62 M:Jnetized Health Benefits of lowered MCL for Sample Year 1988 rnAPI'ER V:

TABLE V-1 Est:inate of Anrrual Per Capita Corrosion Damage V-30 LIST OF FIGURES

PAGE

CHAPI'ER III:

FIGURE III-1 Mllti-organ Irrpa.cts of Lead's Effects on thc9 Heme Pool III-2

FIGURE III-2 Surme.:r:y of Io,,,est Observed Effect Levels for Key Lead-induced Health Effects in Children III-3

FIGURE III-3 Relationship of Blood Lead Level to Weight in Children Aged Oto 7 III-24

FIGURE III-4 Relationship of Blood Lead Level to Height in Children Aged Oto 7 III-25

FIGURE III-5 Flar, Diagram of Medical Protocols for Children with Blood Lead Levels above 2s·ug/dl III-51

CHAPI'ER N:

FIGURE N-1 Adjuste:l Rates of Death and Heart Attacks N-37 versus Blood Pressure: Framingham Data

CHAPI'ER V:

FIGURE V-1 1962 U.S. Geological Survey of Water V-17

CHAPTER I INTRODUCTION

This chapter, which provides an overview of the report, consi~ts of three parts: a background of the regulation of lead under the Safe Drinking Water Act, a summary of this analysis of the ben~fits that could result from a reduction in the amount of lead permitted in U.S. drinking water, and a summary of a case study of the costs and benefits of reducing lead levels in drinking water in the City of Boston.

I.A. Background

The Safe Drinking Water Act {SDWA), passed by the U.S.

Congress in 1974, requires the U.S. Environmental Protection Agency {EPA) to protect public health by setting drinking water standards for public water supplies.* Two levels of protection

are described in the SDWA. Primary drinking water regulations, applicable only to public water systems, control contamination that may have an adverse effect on human health by setting

either a maximum contaminant level {MCL) or a treatment technique requirement. Secondary drinking water standards are non-enforce­ able recommendations concerning the aesthetic quality of drinking water, e.g., taste or smell. The National Primary Drinking Water Reg·ulations {NPDWR)

were first promulgated at the end of 1975. EPA revises those regulations by setting maximum CQntaminant level goals {MCLGs)

* Defined in the Act as water systems serving 25 or more people or having at least 15 service connections. I-2 and related MCLs. MCLGs are non-enforceable health-based goals,. intended to protect against known or anticipated adverse health effects, with an adequate margin of safety. MCLs are enforceable limits, to be set as close as feasible to the MCLG; feasibility includes cost and technological constraints. MCLs are proposed at the same time as the MCLGs.

On November 13, 1985, EPA proposed National Primary Drinking

Water Regulations (NPDWR) to set MCLGs for 28 synthetic organic chemicals, 11 inorganic chemicals, and 4 microbiological parameters in drinking water; these substances are listed in Table I-1. The proposed MCLGs for probable human carcinogens were set at zero, and

MCLGs for other substances were based upon chronic toxicity and other data.

Lead is included among the inorganic substances proposed for regulation in the NPDWR. The current MCL for lead is 50 micrograms of lead per liter of drinking water (ug/1);* the proposed MCLG is

20 ug/1.

The 1986 Amendments to the Safe Drinking Water Act contain a provision banning the use of materials containing lead in public water supplies and in residences connected to public water supplies.

States have until June 1988 to begin enforcing this ban.

I.B. Summary of Report This analysis estimates some of the benefits that could result from reducing exposure to lead in community drinking water supplies.

* This is equivalent to and can be stated alternatively as 0.05 milligrams per liter (mg/1), 0.05 micrograms/gram (ug/g), or 50 parts per billion (ppb). I-3

TABLE I-1. Substances Included in the 1985 Proposed National Primary Drinking Water Regulations (Maximum Contaminant Level Goals)

A. Synthetic Organic Chemicals 1. Acrylamide 13. Ethylene.dibranide 2. Alachlor 14. Heptachlor and Heptachlor 3. Aldicarb, Aldicarb sulfoxide and epoxide Aldicarb sulfone 15. Lindane 4. carbofuran 16. Methoxychlor 5. Chlordane 17. Monochlorobenzene 6. Dibranochloropropane 18. Polychlorinated biphenyls 7. o-,m-Dichlorobenzene 19. Pentachlorophenol 8. cis- and trans-1,2 Dichloroethylenes 20. Styrene . 9. 1,2-Dichloropropane 21. Toluene 10. 2,4-D 22. Toxaphene 11. Epichlorohydrin 23. 2,4,5-TP 12. Ethylbenzene 24. Xylene

B. Inorganic Chemicals 1. Arsenic 2. Asbestos 3. Barium 4. Cadmium 5. Chranium 6. Copper 7. Lead 8. Mercury 9. Nitrate and Nitrite 10. Selenium c. Microbiological~arameters 1. Total Coliform Bacteria 2. Turbidity 3. Giardia 4. Pathogenic Viruses / I-4

These benefits are probably much greater than those attributable to just reducing the MCL for lead, but do reflect benefits attainable with reduced exposure to lead through changes in the MCL coupled with changes in EPA's monitoring requirements or other efforts to reduce exposure to lead from drinking water.

There are two primary categories of benefits evaluated in this paper: the public health benefits of reduced lead exposure (Chap­ ters II and IV) and reduced materials damages (Chapter V) relating to the phenomenon of lead's presence in drinking water as a cor- rosion by-product. In addition, because the calculation of health benefits depends on the extent of human exposure, another chapter

(Chapter II) presents the available data on the occurrence of lead in public water supplies, and presents estimates of the population exposed to drinking water exceeding the proposed MCLG of 20 ug/1.

In assessing the benefits of the proposed reduced lead standard, this analysis assumes that EPA will act to reduce lead levels in tap water, as well as to maintain the current high quality of water leaving the treatment plant. It also relies upon and is sensitive to assumptions about drinking water use and consumption patterns.

This analysis estimates the annual benefits for one sample year, 1988, of lowering the amount of lead permitted from 50 ug/1 to 20 ug/1. That one year was chosen because environmental lead levels will have stabilized following EPA's 1984 phasedown of lead in gasoline.*

* Specifically, this analysis measures effects given the condi­ tions on January 1, 1988, when EPA's proposed ban on leaded gasoline will not yet have taken effect. However, even if EPA promulgates that ban, the estimates in this report will not change significantly. I-5

For comparability, all monetary values are expressed in constant 1985 dollars.* The population baseline is the 219.2 million people served by community water systems.

I.B.l. The Occurrence of Lead in Public Drinking Water

Lead occurs in drinking water primarily as a corrosion by­ product; its sources are the materials used in the distribution and residential plumbing systems (cf sources as diverse as the

US-EPA Air Quality Criteria Document for Lead, 1986; Craun and

McCabe, 1975; Kuch and Wagner, 1983; Department of the Environment,

1977; etc}. Water leaving the water treatment plant is usually relatively lead-free. However, pipes and solder containing lead are corroded by water, and lead levels at the user tap are often much higher than those found at the treatment plant. While the presence of lead service pipes is relatively restricted geographi­ cally in the United States, the use of lead solder {and flux} is ubiquitous. And the combination of copper pipes with solder containing lead found in most residences can result in high lead levels** in first drawn water that has been in contact with the pipe for a period of time -- levels exceeding the current MCL, even with fairly non-corrosive waters {e.g., Nielson, 1976). In particular, newly-installed solder is easily dissolved, and people living in new housing, or in older housing but with new

* The 1986 Economic Report of the President to Congress (Table B-4).

** This results from galvanic corrosion, which is the corrosion that occurs when 2 metals, with different electro-chemical potential, are in the same environment. I-6 plumbing, are especially at risk of high levels of lead in the drinking water (Sharrett et al., 1982a: Murrell, 1985). Lead concentrations in fully flushed water typical of disfribution system water, even under corrosive conditions and with new solder, are generally below 50 ug/1 and usually below 20 ug/1.

Because lead occurs generally as a corrosion by-product in

U.S. community water supplies, levels .in fully-flusned water and in distribution water are typically low. Exposure to lead, however, is from tap water that can contain significant amounts of lead.

The estimate of the occurrence of lead in drinking water, therefore, is based upon data collected and analyzed for EPA's Office of

Drinking Water in 1979-81. These data portray lead levels partly flushed (30 seconds), kitchen tap samples collected by the Culligan water-softening company.* J. Patterson of the Illinois Institute of Technology analyzed the data. Current evidence indicates that these samples are more representative of consumed water than are the fully-flushed samples taken in compliance with EPA's monitoring regulations. The Ctilligan data indicate that 16 percent of partly flushed water samples could exceed an MCL of 20 ug/1 at the kitchen tap. The findings from this data source are consistent with other analyses of the occurrence of lead in tap water and with studies of lead leaching rates in corrosive and non-corrosive waters.

To this must be added the inhabitants of housing built within the past 24 months and that have plumbing materials containing lead.

* The use of company names and the presentation of related data does not 9onstitute endorsement of their services. I-7

Many studies have shown that new solder can release significant amounts of lead into water, even exceeding the current MCL of 50 ug/1 (e.g., Sharrett et al., 1982a; AWWA-DVGW Cooperative Research

Report, 1985). While corrosive waters have the highest lead levels, relatively non-corrosive waters can also leach significant amounts of lead. The highest lead contamination levels occur with the newest solder (i.e., during the first 24 months following installation), but those levels decline and are generally not elevated beyond five years (e.g., Sharrett et al., 1982a;

Lassovszky, 1984).

There were 1.7 million new housing starts and permits in the

United States during 1983 and 1.8 million in 1984.* Construction data show that housing typically takes six months to a year from permit to potential occupancy, so there are currently about 3.5 million new housing units (i.e.,< 24 months). The Statistical

Abstract of the United States (1985; Table 58) indicates that the average household contains 2.73 individuals. Multiplied tog~ther, a total of 9.6 million people currently live in new housing.

However, not all of these people are served by community water supplies: of the current (1985) U.S. residential population of over 240 million, 219.2 million anf served by community water systems and this analysis only addresses that population. In addition, data from the plumbing supply industry show that about

* Survey of Current Business, U.S. Department of Commerce - Bureau of Economic Analysis, 1985; Table on New Housing Construction. I-8

8 percent* of new plumbing is plastic, so 92 percent of the population has metal pipes. Therefore, the number of people at risk of high lead levels from new solder in new housing is:

219 mil 9.6 mil x 240 x .92 = 8.1 million.

To calculate the risk to inhabitants of older housing, subtract the number in new housing (8.8 million)** from the total served by community water systems (219.2 million); that indicates that 210.4 million people live in older homes. Based upon the Culligan data, 16 percent of them (33.7 million) are at-risk of high lead levels from partly flushed water at the kitchen tap. Combining the data from Culligan on lead levels in older housing with the new housing exposure estimates indicated that 41.8 million people using public water supplies currently may be exposed to some water that exceeds the proposed MCL of 20 ug/1; we round this to 42 million.

This may be a low estimate

0 because it does not include the potential exposure of occupants in housing built within the past 2-5 years (who also probably remain at greater risk of elevated lead levels);***

0 because we have not included those who, while living in older housing, have recently had major plumbing repairs and so are also at risk of the potentially high lead levels associated with newly installed solder;

* This is the average of claims by the Plastic Pipe Institute presented in Mruk (1984) and of the Copper Development Association presented in Anderson (1984).

219 mil ** Derived: 9.6 mil x 240 = 8.8 million people.

*** Inhabitants of 2-5 year old housing are not included in this analysis because it was not possible to eliminate them from the base and thus avoid double-counting. I-9

0 because the Culligan data represent water that is harder than average, whereas high lead levels are often found with soft waters; and

0 because the data used are for partially flushed samples, while some people (especially children) may consume water that is closer to first-flush or standing samples (which is more likely to contain higher concentrations of lead).* In addition, we have not included any data from the estimated

60 million people served wholly or in part by private and non­ community water supplies.

There are uncertainties, however, concerning actual patterns of drinking water use and the extent of plastic piping in new construction that could reduce the estimate. Early enforcement of the Safe Drinking Water Act ban on the use of materials containing lead in public water supplies, enforceable in June 1988, could also decrease exposure to lead from drinking water. The assumptions on the relationship between water lead levels and blood lead levels are taken from the draft (EPA) Water Criteria

Document for Lead (1985), which is based upon the recommendations in the Air Quality Criteria Document for Lead (US-EPA, 1986). Those documents assume a linear relationship, at least at the lower blood-lead levels typical of the United States, with dif­ ferent constants for children and adults relating first-flush water lead levels to blood lead concentrations. Those formulae are:

* Water standing in pipes has a greater opportunity for lead to leach into it and, therefore, is more likely to contain higher lead levels. Many of the factors affecting lead levels in drinking water are discussed in Chapters II and V of this report. I-10

(for children}t PbB* = 0.16** x intake of lead from water (for adults} PbB* = 0.06** x intake of lead from water.

Alt~rnative assumptions (e.g., those reasonably derived from the results of Richards and Moore, 1982 and 1984} could imply that exposure and consequently benefits -- may be underestimated, possibly by several factors.

The estimates of the health benefits associated with this pro­ posed rule rely upon data on the distribution of blood lead levels in children and adults collected as part of the Second {U.S.}

National Health and Nutrition Examination Survey (NHANES II}, a

10,000 person representative sample of the U.S. non-institutional­ ized population, aged 6 months to 74 years. That data base is available from the {U.S.} National Center for Health Statistics and analyses of lead-related data from it have been published before (e.g., Annest et al., 1982 and 1983; Mahaffey et al.,

1982; Pirkle and Annest, 1984}. This analysis· uses both point estimates and ranges of blood lead levels associated with specific health outcomes. Other EPA analyses {e.g., US-EPA, 1986a} use ranges exclusively. Both approaches are supported by the available data.

I.B.2. Benefits of Reducing Children's Exposure to Lead Elevated blood-lead levels have long been associated with neurotoxicological effects and many other pathological phenomena:

* PbB = blood lead level ** These constants have a unit of micrograms-of-lead/deciliter-of­ blood per microgram-of-lead-in-water/day, or ug/dl per ug/day. t This formula was derived from Ryu, 1983. An alternative estimate from the data in that paper suggests a coefficient of about 0.4. I-11 an article on lead's neurotoxicity was published as early as 1839, on anemia in the early 1930s, on kidney damage in 1862, and on impaired reproductive function in 1860. As noted in the Air

Quality Criteria Document for Lead (US-EPA, 1986), from an his­ torical perspective, lead exposure levels considered acceptable for either occupationally-exposed persons or the general popula­ tion have been revised downward steadily as more sophisticated bio-medical techniques have shown formerly-unrecognized biological effects, and as concern has increased regarding the medical and social significance of such effects. In the most recent downward revision of maximum safe levels for children, the Centers for Disease Control {CDC) lowered its definition of lead toxicity to

25 micrograms of lead per deciliter of blood (ug/dl, the standard measure of blood lead level) and 35 ug/dl of free erythrocyte protoporphyrin (FEP). • As evaluated in the Criteria Document

(1986), the present literature shows biological effects as low as 10 ug/dl (for heme biosynthesis) or 15 ug/dl (for certain renal system effects and neurological alterations); indeed, a threshold has not yet bc3en found for some effects (e.g., elevated levels of a potential n13urotoxin* or stature effects, Angle et al., 1982; Schwartz et al., 1986). There is no convincing evidence that lead has any beneficial biological effect in humans (Expert Committee on Trace Metal

Essentiality, 1983; and included in the Criteria Document, 1986).

Elevated blood-lead levels have been linked to a wide range of health effects, with particular concern focusing on young

* ALA, or aminolevulinic acid. I-12 children. These effects range from relatively subtle changes in biochemical measurements at 10 ug/dl ano below, to severe retardation and even death at very high levels (80-100 ug/dl).

Lead can interfere with blood-forming processes, vitamin D metabolism, kidney function, neurological processes and repro­ ductive functions in both males and females. In addition, the negative impact of lead on cognitive performance (as measured by

IQ tests, performance in school, and other means) is generally accepted at moderate-to-high blood-lead levels (30 to 40 ug/dl and above), and several studies also provide evidence for possible attentional and IQ deficits, for instance, at levels as low as

10-15 ug/1. Changes in electroencephalogram readings, as another example, have also been observed at these low levels. For many subtle effects, the data may represent the limits of detectability of biochemical or other changes, and not necessarily actual thresholds for effects. For children's health effects, two categories of benefits were estimated monetarily: 1) the avoidance of costs for medical care for children exceeding the lead toxicity level set by the

Centers for Disease Control (i.e., 25 ug/dl, when combined with

FEP levels of> 35 ug/dl) and 2) the averting of costs due to lead-induced cognitive effects. Two alternative methods for valuing the potential cognitive damage resulting from exposure to lead were developed. The first of these two alternatives involves assessing the costs of compensatory education to address some of the manifestations of the cognitive damage caused by lead as a I-13 proxy measure for the damage itself. The second relates to one specific indication of that cognitive damage -- potential IQ point loss, and includes a calculation of decreased expected future earnings as a function of IQ point decrement. These esti­ mates neither include many major categories of pathophysiological effects (e.g., renal damage), nor do either the medical costs or the compensatory education costs consider any lasting damage not reversed by medical treatment or compensatory education. These estimates also attribute few benefits to reducing lead levels in children whose blood lead levels would be below 25 ug/dl even in the absence of this proposed rule.

The estimate of reductions in medical care expenses rely upon published recommendations (Piomelli et al., 1984) for follow-up testing and treatment for children with blood lead levels above 25 ug/dl. The costs of such medical services and treatment were estimated at about $950 per child over 25 ug/dl (1985 dollars). This average reflects both lower costs for most of these children and much higher costs for the smaller subset requiring chelation therapy.

The estimates for compensatory education assumed three years of part-time compensatory education (de la Burde and Choate, 1972 and 1975) for 20 percent of the children above 25 ug/dl, averaging about $2,800 (1~85 dollars) per child above that blood lead level based upon data from the u.s. Department of Education (Kakalik et al., 1981).

There is extensive literature examining the relationships between IQ, educational levels attained, demographic variables and I-14 earnings (ICF, 1984). The results of that literature were used to estimate the effect of IQ point losses that can occur as a part of the cognitive damage caused by lead exposure upon expected future earnings: one IQ point can directly and indirectly affect earnings by 0.9 percent. The studies of cognitive damage presented in the Air Quality Criteria Document for Lead (U.S. EPA, 1986) show evidence that blood lead levels of 15-30 ug/dl can be associated with IQ losses of 1-2 points, blood lead levels of 30-50 ug/dl can be associated with IQ losses of 4 points, and over 50'ug/dl of blood lead can cor~elate with losses of 5 points. Data from the

Census Bureau on expected future lifetime earnings, deferred for 20 years* at a 5 percent real discount rate and then annualized, yield estimated benefits of avoided damage fr.om reduced exposure to lead.

This alternative method for valuing some of lead's cognitive damage indicated that society could save $1,040 per child brought below 15 ug/dl; $2,600 per child brought below 30 ug/dl; and $2,850 per child brought b~low 50 ug/dl by reducing lead in drinking water

(1985 dollars).

In sum, this analysis indicates that the proposed rule could produce benefits of $27.6 million annually in avoided medical expenses; $81.2 million per year in reduced compensatory education costs; and $268.1 million per year in increased lifetime earnings,

* These costs are deferred because those suffering the effects are children and will not enter the work force for up to 20 years. Obviously, using the largest deferral period (20 years) reduces the value of the benefit and reduces the benefit estimate, whereas 8- or 10-year-old children may begin working within 8 years and so would have a much shorter deferral period. This biases the estimates downward slightly. I-15 based upon sample year 1988: these estimates are in 1985 dollars. Note that compensatory education and affected earnings are alter­ native methods for valuing aspects of the cognitive damage caus-ed by lead.* In addition, benefits potentially derived by decreasing the incidence of two other categories of h~alth effects (lead's adverse effect upon children's growth and fetal effects) were not estimated in dollar terms. Assuming that pregnant women are distributed proportionately throughout the country, data from the Census Bureau** on birth rates and demographic distributions indicate that 24 per­ cent of the total population is women of child-bearing age (15-44) and that the birth ratie is 67.4 births per 1,000 women aged 15-44.

Therefore,

41.8 million x 24% x 67.4 per thousand= 680,000.

It is estimated that this proposed rule could prevent 680,000 fetuses from being exposed to elevated lead levels. The fetal effects are particularly important, because several recent studies have shown that lead exposure within the normal range ( 6-2·0 ug/1) can be associated with various negative pregnancy outcomes (such as early membrane rupture and even miscarriages, e.g., Moore, 1982: Wibberly et al., 1977), and with low birth weight, inhibited post­ natal growth and development (e.g., Bornschein, 1986: Bellinger, 1985 and 1986: ~ietrich et al., 1986). In addition, this proposed

rule could prevent 82,000 children from risk of growth effects.

* This also biases the results downward because there is a strong rationale for considering these effects as additive.

** Statistical Abstracts (1986), Tables 27 and 82. I-16

I.B.3. Blood-Pressure-Related Benefits and Other Adult Health Effects Lead has long been associated with elevated blood pressure, but until recently most of the studies have focused only on hyperten­ sion and relatively high lead levels typically found only in those occupationally exposed to lead. Several recent studies, however,

(e.g., Pirkle et al., 1985; Harlan et al., 1985; Pocock, 1984 and

1985), have found a continuous relationship between blood lead and blood pressure. These studies provide evidence for a small (com­ pared to other risk factors) but robust relationship after control­ ling for numerous other factors known to be associated with blood pressure. Experimental animal studies in several species of rats and pigeons also provide evidence of a relationship between moderate blood-lead levels and increases in blood pressure.

To calculate these benefits, logistic regression equations were used to predict how reducing exposure to lead in drinking water would affect the number of hypertensives in the u.s. popu­ lation. These estimates cover only males aged 40 to 59, because the effect of lead on blood pressure appears to be stronger for men and because the correlation between blood pressure and age is much smaller in this age range, reducing the potential for confounding due to the correlation between blood lead and age. The estimates rely upon 1) site-adjusted coefficients from analyses of the NHANES

II data relating blood lead levels to increases in blood pressure* and 2) coefficients relating blood pressure increases to more serious

* The specific coefficients and the basis for their derivation are described in the Addendum to the Criteria Document, 1986, which is included in volume 1 of that publication~ I-17

cardiovascular disease outcomes, based on data from the Framingham

Study (McGee et al., 1976) and Pooling Project (1976), as confirmed

by Levy et al., 1984.

Levy has demonstrated that the risk coefficients from the

Framingham Heart Study, when coupled with the observed reductions

.....in blood pressure, smoking, and cholesterol in the U.S. population

during the 1970s, correctly predicts the observed reductions in

cardiovascular mortality in the overall population during that

decade. The Pooling Project showed that the Framingham coefficients

adequately predicted cardiovascular outcomes (such as strokes and

heart attacks) in the other five large prospective heart studies performed in the u.s. Therefore, while caution is clearly warranted in view of the limited data on the effect of lowering blood lead

levels on blood pressure, use of the regression coefficients from the Framingham Study provide a reasonable basis by which to predict

potential changes in cardiovascular outcomes associated with blood

pressure changes due to decreased lead exposure.

Based upon this information, reducing exposure to lead from

drinking water in 1988 could reduce the number of male hypertensives

(aged 40 to 59) by 130,000. Using estimates of the costs of medical

care, medication, and lost wages, such a reduction in hypertension

incidence would yield a value of $250 per year per case avoided (1985

dollars).

These estimates o_f how blood pressure reductions would affect

the incidences of various cardiovascular diseases were based on

projections of changes in blood pressure as a result of the proposed

rule and estimates of the relationships between blood pressure and I-18 heart attacks, strokes, and deaths from all causes. As noted earlier, the latter estimates were derived from several large epidemiological studies, primarily the Framingham study. However, because those studies included very few nonwhites, the estimates were further restricted to white males, aged 40 to 59. Thus, the benefits estimates do not include middle-aged, nonwhite males.

The basis of most of the medical costs are the'cost-of-illness estimates presented in Hartunian et al., 1981, which were adjusted in three ways to reflect current conditions. First, we inflated them to 1985 dollars using data from the 1986 Economic Report of the President to Congress. Second, we adjusted the costs to reflect changes and improvements in medical treatment, including the trip­ ling in the incidence rate of coronary bypass operations that occurred between 1975 and 1982. Third, Hartunian used a 6 percent real discount rate to present-value future expenditures, while this analysis uses a 10 percent real discount rate.

The value 0£ reductions in heart attacks and strokes was based on the cost of medical care and lost wages for nonfatal cases.

Expected fatalities from heart attacks and strokes were included in the estimate of deaths from all causes. That procedure yielded benefits of $65,000 per heart attack avoided and $48,000 per stroke avoided (1985 dollars) for the 240 heart attacks and 80 strokes estimated to likely be avoided in 1988 because of this proposed rule. It is important to note that these estimates do not account for any reductions in the quality of life for the victims of heart attacks and strokes (e.g., the partial paralysis that afflicts many stroke victims). I-19

Valuing reductions in the risk of death is difficult and con­

troversial, with a wide range of estimates in the literature. EPA's policy guidelines (U.S~ EPA, 1984c), for example, suggest a

range of $400,000 to $7 million per statistical life saved. Using

$1 million per case, the benefits of reduced mortality dominate our estimates of total blood-pressure-related benefits~ these total

$240 million in 1988 for the 240 deaths estimated as likely to be avoided in that year. Altogether, the monetized benefits of reducing adult male exposure to lead in drinking water are estimated to total $291.9 million per year (in 1985 dollars), using 1988 as a sample year.

In addition, because lead crosses the placental barrier and is a fetotoxin, pregnant women exposed to lead are at risk of compli­ cations in their pregnancies and da~age to the fetus. (Fetal ,, I effects are discussed above, under 6hildren's health effects.)

While we have not monetizeo any of these reproductive effects, as noted above, 680,000 pregnant women per year probably receive I' water that exceeds the proposed standard of 20 ug/1, and would benefit from the proposed rule. Lead-induced effects on male reproductive functions have also been discussed in the scientific literature but are not included in this report.

I.B.4. Benefits of Reduced Materi~s Damage ! A third category of monetized 1benefits relates to the phenom- enon of lead's presence in drinking\water: it is a product of the corrosive action of water upon the materials of the distribution ' and residential plumbing. For tne most part, therefore, treatment I-20 processes used to reduce high levels of lead in drinking water are the same as treatment processes used to reduce the corrosion poten­ tial of the water. Reducing corrosion damage will produce substan­ tial benefits to water utilities, their rate-paying customers, and building owners. Published estimates of the costs of corrosion damage range from $12 to $46 per person per year (1985 dollars), and are sum­ marized in Table I-2. Estimates of the costs that can be avoided by corrosion control measures range from 20-50 percent of total damage. The point estimate of avoidable corrosion costs (i.e., the benefits of corrosion control) is $8.50 per capita annually (1985 dollars). For comparison, estimates of average corrosion treatment costs range from under $1 per person per year (based upon the experience in 18 cities currently treating their highly corrosive waters) to about $5 per person per year (based upon the highest treatment costs presented in the ODW cost report).* As a point estimate, we assumed per capita annual treatment costs of $3.80

(1985 dollars). Estimates of the extent of corrosive water also vary. A commonly accepted profile is that developed by the U.S. Geological

Survey in the early 1960s, which identified the Northeast,

Southeast, and Northwest sections of the country as having the softest and most corrosive wat~rs· (Durfor and Becker, 1964a and

* The range, however, is quite wide and highly sensitive to system size. These represent average costs. In some very small systems (i.e., serving 25-100 people),_ costs may be many times higher. I-21

TABLE I-2. Estimates of Annual Per Capita Corrosion Damage (1985 dollars) Corrosion Estimated Annual Corrosion Damage Damage ( >er capita) Avoidable Annual Per Through Capita Benefits Studies Distribution Residential Total water of Corrosion Asst.mtptions/Notes Systems Treatment Control Kennedy Engineers Assumed 30% potential reductiqn (1973) $5.57 - $16.71* 30%* $5.01* in corrosion damage and that dis- tribution costs were one-third of total costs. Hudson & Gilcreas They did not include increased (1976) $8.68* - $26.04* 50% $13.02* operating costs. Per capita estimate assumes 200 million people are served by public water systems. Assumed that distri- bution costs were one-third of total costs. Kennedy Engineers They calculated $6.17 per capita (1978) - $30.87* $46.30* 20% $9.26* in savings to residence owners. Assumed residential costs were two-thirds of total costs. Bennett et al. Assumed that 200 million people (1979) $9~40 - $28.20* 20% $5.64* are served by public water systems (cited in Ryder, and that distribution costs were 1980) one-third of total costs. Energy & Environ- This 1s an admitted underestimate: mental Analysis $3.98 $7.97 $11.95 38% $4.54 it includes only damage to pipes (1979) (not damage to w~ter heaters, increased oneratina' costs, etc.) Ryder (1980) $1.17 $22.19 $23.36 25% $5.84 Ryder ascribed 95% of corrosion damage to private owners. Kinneyer & Logsdon -- ~23.60-X ~35.40-X 40% ~14.16-X Assumed residential costs were (1983) two-thirds of total damage. AVERAGE $8. 21 W/OUT EEA $8. 82

* These estimates have been calculated by the authors of this paper. Assumptions are noted above. I-22

1964b). The combined populations of those states are 67.7 million people (1980 census). Assuming that these areas are served proportionately by community water systems,* 61.8 million people would benefit from actions to reduce the corrosivity of their water.

That figure, multiplied by $8.50 per person, yields annual benefits from reduced corrosivity of $525.3 million (1985 dollars).

I.B.5. Summary of Annual Benefits of Reduced Lead i·n Drinking Water This analysis of the benefits of reducing exposure to lead in drinking water indicates that the monetized annual benefits could range from $926.0 to $1,112.9 million (1985 dollars) for sample year 1988. In addition, there are numerous health benefits of reduced exposure to lead that are not monetized. The annual monetized benefits are summarized in Table I-3, and the non­ monetized benefits are presented in Table I-4. Based upon the latest cost estimates used by the Office of

Drinking Water** the projected benefits exceed the costs by about

4:1. Expressed differently, lowering the MCL to 20 ug/1 could produce annual net benefits of about $800 million in 1988. It should be emphasized that considerable uncertainty is associated with these estimated benefits, uncertainties derived both from the current state of knowledge concerning lead health effects and the valuation of avoiding such effects. Other

* Of the total population of about 240 million, 219.2 million people are served by community water systems.

** These calculations use preliminary EPA Office of Drinking Water cost estimates. Costs and net benefits will be discussed more extensively in other documents associated with this proposed rulemaking. I-23

TABLE I-3. Surrmary of Estimated Annual Monetized Benefits of Reducing Exposure to Lead from 50 ug/1 to 20 ug/1 (1985 dollars) for Sample Year 1988

Estimated population exposed to drinking 42 million* water exceeding proposed MCL

Children's health benefits -reduced medical costs $27.6 million -reduced costs of cognitive damage Method 1 - canpensatory education $81.2 million Method 2 - decreased future earnings $268 .1 million

'IOTAL: Method 1 $108. 8 million Method 2 $295.7 million

Adult health benefits (males only) -reduced hypertension savings $32.5 million (males, aged 40-59) -savings from fewer heart attacks. $15.6 million (white males, aged 40-59) -savings from fewer strokes $3.8 million (white males, aged 40-59) -savings from fewer deaths $240.0 million (white males, aged 40-59) TOTAL: $291.9 million Materials benefits -benefits of reduced corrosion damage $525.3 million

TOTAL ANNUAL IDNETIZED BENEFITS -Method 1 - using canpensatory education $926.0 million -Method 2 - using decreased future earnings $1,112.9 million ESTIMATED ANNUAL COSTS $230.0 million NET ANNUAL MONETIZED BENEFITS about $800 million

* Total population served by corrmunity water systems: 219 million I-24

TABLE I-4. Surrrrary of Estinated Annual Non-rronetized Benefits of Reducing Exposure to Lead fra:n 50 ug/1 to 20 ug/1 for Sanple Year 1988 Reduct.ions in Nurrbers of Pecple at Risk

Estinated J,X?PUlation exposed to drinking 42 million* water exceeding prcposed ~ Ori.ldren's health benefits

- cnildren ra;iuiring rredical treatnent 29,000

- loss of 1-2 IQ points 230,000 4 IQ points 11,000 5 IQ points 100 - dlildren ra;iuiring canpensatory education 29,000

- dlildren at risk of stature decrerrent 82,000

- fetuses at risk 680,000

- increased risk of herratological effects 82,400

Adult health benefits

-cases of hypertension 130,000 {males, aged 40-59)

-heart attacks 240 {white males, aged 40-59)

-strd

-deaths 240 (white nales, aged 40-59)

-(reduced risk to pregnant wcmen 680,000) ((waren, aged 15-44) ) (sane as fetuses )

* Total population served 1::¥ canrnmi-cy -water systems: 219 million I-25 analogous efforts to estimate benefits associated with reducing lead in drinking water may be useful in helping to judge how reasonable these present benefit estimates are.

I.e. Boston Case Study In the spring of 1986, Jonathan Jacobson analyzed the incre­ mental costs and benefits to the City of Boston of reducing the lead MCL from 50 ug/1 to 10 ug/1 (Jacobson, 1986).* That analysis, carried out as a masters thesis project at Harvard University, focused on Boston as a city with high potential for increased lead exposure via drinking water.

Boston's water is highly corrosive: it is relatively acidic

(pH= 6.7) and soft (14 mg/1 as CaCO3), and has low alkalinity (Karalekas et al., 1975). Boston also has a large percentage of lead pipes in service. During the 1970s, several studies found high lead levels in Boston's drinking water (e.g., Karalekas et al., 1975; and several internal studies conducted by the

Massachusetts Water Resources Authority and the [Boston]

Metropolitan District Commission). To reduce the high lead levels, Boston began corrosion control treatment. Monitoring performed by EPA's Region I from

1975 to 1981 indicated that lead, iron and copper levels dropped significantly (Karalekas et al., 1982). However, while lead concentrations generally decreased to below the current MCL of

50 ug/1, additional treatment will be necessary to comply with a lowered MCL.

* He assumed a lowered MCL of 10 ug/1 because that is the feasibility limit for current treatment and technology. I-26

Jacobson analyzed the incremental annual costs and benefit~ of additional efforts by Boston to control further the corrosivity of its water, using sample years 1988 and 1992, and estimating all costs in 1985 dollars. His analysis assumed the following:

0 compliance will be measured by a standing "grab" sample, that is, a sample taken immediately after turning on the faucet at any random time during the day after an unknown period of standing;

0 it wiil be impossible for every tap to meet the lowered MCL, even using state-of-the-art treatment, and so samples should be averaged; and

0 that, while the effectiveness of specific treatment procedures varies in not-yet-well-understood ways when actually used in the field, corrosion control is ulti­ mately feasible with current state-of-the-art methods.

Jacobson, using data from EPA Region I and the Massachusetts water Resources Authority, calculated the benefits of additional

2-stage treatment for Boston's water: further raising the pH (to reduce the acidity of the water) and installing several pumping stations to maintain a consistent concentration of sodium hydroxide throughout the system. Jacobson used the same categories of monetized health benefits*

as those described in this EPA analysis, except that he did

not include the estimates of cognitive damage associated with

decreased future earnings. His estimates of materials benefits

rely only upon the Kennedy Engineers (1978) study and information

in the American Water Works Association Corrosion Manual (AWWA-DVGW,

1985).

* For blood-pressure-related estimates, however, he used the non­ site-adjusted coefficients from the NHANES II contained in Pirkle et al., 1985. I-27

His results, summarized in pages 36-39 of his study, indicate

incremental costs of $700,000 per year (using sample year 1988) and incremental annual benefits of $7.9 million (including estimated health benefits of $6.9 million and materials benefits of

$950,000; based upon sample year 1988). This yields estimated net annual benefits of $7.2 million and a benefit to cost ratio of 11:1 (compared to the estimate of 4:1 in this analysis).

It is, however, unclear whether or how these results can be extrapolated to other U.S. water systems and cities, and therefore, to this proposed rule.

CHAPTER II

OCCURRENCE OF LEAD IN DRINKING WATER

Lead can contaminate drinking water through several pathways.

It can result from naturally present lead in the source water, it can result from contamination of the water supply, or it can result as a by-product of corrosive water.* In the last case, the sources of the lead are the pipes, plumbing fixtures, flux and solder of the distribution system and within private residences or other buildings. Most contamination of drinking water with lead results from the corrosion of materials containing lead.

Section A of this chapter discusses the sources and factors affecting the contamination of drinking water by lead.** Section B of this chapter discusses the available data on the occurrence of lead in community drinking water supplies. Because lead occurs in drinking water primarily as a corrosion by-product, contamination

* Corrosion is the deterioration of a substance or its proper­ ties due to a reaction with its environment. In this paper, the "substance" that deteriorates is the pipe whether made of metal, asbestos-cement, cement, or plastic and the flux and solder joining the pipes, and the "environment" is water. That is, we are concerned with internal corrosion. (Pipes and other water treatment equipment can also corrode externally.)

** Chapter V also contains a discussion of the relationship between corrosive water and lead in drinking water, but from the perspective of potential corrective action. II-2 levels are higher in tap water in homes* than in the water leaving the water treatment plant or flowing through the water mains.

Section C contains EPA's estimates of exposure to lead in the drinking water provided by community water systems in the United States.

The exposure estimate presented in this chapter serves as the basis for calculating the benefits from a potential reduction in exposure due to a lowering of the Maximum Contaminant Level

(MCL) for lead from the current 50 micrograms of lead per liter of water (ug/1) to 20 ug/1.**

II.A. sources of Lead in Drinking water

The principal source of lead in ambient surface water is anthropogenic lead particulates from the atmosphere, which come mostly from the combustion of. leaded gasoline (e.g., Laxen and

Harrison, 1977; Trefry et al., 1985) and, to a lesser extent, the smelting of ores and the combustion of fossil fuels. Evidence indicates that much of the lead in surface waters will end up in sedimentary deposits (Laxen and Harrison, 1977), and most of the

* Lead contamination of drinking water as a result of corrosion also occurs in commercial buildings, schools, etc. There is less information on the factors that determine lead levels· in these buildings, however, than on the use patterns and materials in private homes. Therefore, this analysis examines only exposure to lead in residential circumstances. Additional research is needed in other areas of exposure, including the work place (factories, office buildings, etc.) and in schools. ** This is equivalent to and can be expressed as 0.020 milligrams per liter (mg/1), 0.020 micrograms per gram (ug/g), or 20 parts per billion (ppb). II-3

lead dissolved in rain will precipitate and be filtered out by the soil. Lead from runoff or fallout will also precipitate and be retained within the soil or sediments.

The source of most lead in ground water is geochemical; that is, the minerals contained in the rocks and soil through which the ground water flows. Concentrations of lead in ground water in the United States are typically under 10 ug/1 (Lovering, 1976) and,

generally, lead solubility is very low in ground water.

The principal sourc~e of lead in drinking water is neither naturally occurring lead in ground water nor anthropogenic lead in

surface water, however. It is the materials of the water supply

and distribution systems and the plumbing in homes and other buildings. (See sources as diverse as Craun and McCabe, 1975;

[U.S.] National Academy of Sciences, 1977; EPA's Lead Technologies and Costs Document, 1984; Kuch and Wagner, 1983; AWWA-DVGW Coopera­ tive Research Report, 1985; EPA's Lead Occurrence Document, 1985;

EPA's Air Quality Criteria Document for Lead, 1986; etc.). The highest concentrations of lead are found where pipes or solder

containing lead are used in combination with corrosive waters, where water has been sitting for many hours, or where there is

newly-installed piping or repaired joints using flux or solder containing lead.

II.A. 1. Variables Affecting Lead Levels in Drinking Water

The lead used in service pipes* or as part of lead/tin solder is relatively resistant to corrosion under simple laboratory

* Service pipes connect the main pipes of the water distribution system to the plumbing contained within the home. II-4 conditions. In addition, it can be protected from corrosion by a thin layer of relatively insoluble film, composed of carbonate compounds of lead or calcium, that forms on the surface of the metal (Patterson and O'Brien, 1979; Schock and Gardels, 1983;

Lassovszky, 1984; etc.). Water, however, is a corrosive substance.

In addition, the combination of copper piping with tin/lead solder found in most residences can result in galvanic corrosion,* which can yield lead levels much greater than expected from the simple corrosion of the water alone.

Other conditions typical in private homes exacerbate the results of galvanic corrosion and can contribute to high lead levels in home tap water.** These include the facts that residen­ tial plumbing materials are often less corrosion resistant and well-protected than those used in distribution systems {AWWA

Committee Report, 1984), that water often remains overnight in household pipes, and that some water used in private homes is heated, greatly increasing its corrosive potential.

II.A.1.a. Key Water Parameters Affecting the Solubility of Lead

All water is corrosive to some degree. However, some quali­

ties of water make it more corrosive for certain materials. The solubility of lead {also called plumbosolvency) is complicated.

* Galvanic corrosion results when two metals, with different electrochemical potential, are in the same environment.

* The maximum equilibrium level is determined by the specific qualities of the water. Because waters in public drinking systems rarely reach equilibrium levels {for purposes of corrosion), non-equilibrium conditions are assumed in the analysis presented in this document. II-5

Probably the master control variable in the solubility of lead is

pH,* although it is closely interrelated with carbonate.** Waters below pH 6.8 are very plumbosolvent, as can be waters with

pH above 10.2 (e.g., Moore, 1973; Schock, 1980; Sheiham and Jackson, 1981; Murrell,, 1985) , but pH does not have a strictly

linear relationship to lead levels in water (e.g., Patterson and

O'Brien, 1979; Pocock, 1980; Schock and Gardels, 1983). In addition, at low or very high carbonate alkalinities,** lead is

soluble throughout the pH range of drinking water (e.g., Department

of the Environment, 1977, Pocock, 1980; Jackson and Sheiham,

1980; Schock, 1980; Sheiham and Jackson, 1981; Kirmeyer and

Logsdon, 1983; Schock and Gardels, 1983; Gregory and Jackson,

1984; AWWA-DVGW Report, 1985). Soft watert is usually plumbosol­

vent, but several studies have shown that very hard water can also be plumbosolvent (e.g., Department of the Environment, 1977;

* The factors of water that inhibit or enhance corrosion are discussed in Chapter v, including measures of those para­ meters. In short, pH is a measure of the concentration of hydrogen ions (H+) in the water, which is important because H+ is one of the major substances that determines how much metal can be corroded electrochemically. ** The carbonate content (measured indirectly by alkalinity and pH, and usually given in units of equivalent calcium carbonate, caco3 ) relates mostly to the presence of dissolved bicarbonate and carbonate ions in the water and enables the formation of a relatively insoluble protective coating on the inside of the pipe, forming a barrier between the water and the materials of the plumbing system. t Water with low levels of calcium and magnesium ions, which can help form a protective coating on the inside of the pipe. Hardness is also expressed as the equivalent quantity of CaC03 (calcium carbonate). II-6

Thomas et al., 1981)., Other properties of the water, including temperature (e.g., Mullen and Ritter, 1980; Britton and Richards,

1981); velocity; treatment with chlorination or chloramination*

(Treweek et al., 1985}; presence of humic substances (Moore, 1973; Samuels and Meranger, 1984}; chloride and nitrate levels

(Oliphant, 1982); and dissolved oxygen or other elements, may also affect plumbosolvency.

II.A.l.b. Lead Solder The use of lead/tin solder, in a tin to lead ratio of 50:50 or 60:40, is ubiquitous in U.S. residential plumbing at present

(e.g., Lead Industries Association, 1982; Chin and Karalekas, 1984; AWWA-DVGW Report, 1985). Lead solder is probably the greatest single contributor to lead contamination of drinking water in this country because of its widespread use and easy solubility. Its easy solubility is caused by the galvanic reaction between the lead/tin solder and the copper pipes that are used most commonly in residential plumbing (Anderson, 1984). Many recent studies have shown that solder containing lead, when used with copper household plumbing, could easily raise lead levels above 50 ug/1, even when in contact with relatively non-corrosive water or within a relatively short period of time (e.g., Wong and Berrang, 1976; Nielsen, 1976; Department of the Environment,

1977; Lyon and Lenihan, 1977; Lovell et al., 1978; Britton and Richards, 1981; Sharrett et al., 1982a and b; Oliphant, 1982 and

* Chemicals used in drinking water disinfection. II-7

1983; Samuels and Meranger, 1984; Birden et al., 1985; Treweek et al., 1985}. Some of these and other investigations (e.g., Depart­ ment of the Environment, 1977} also found that lead solder alone could produce lead contamination levels that were as high or higher than those in wholly lead-plumbed houses. Indeed, under some conditions, because of the galvanic action they can be much higher (Oliphant; 1982 and 1983; Murrell, 1985}. Data summarized in the AWWA-DVGW Report (1985}, Table 4-19 and elsewhere, show that even without new solder, the galvanic reaction in relatively non-corrosive waters (pH 7.5-8.5, alkalinity> 100 mg/1 as caco3} can produce lead levels at the tap of 160-250 ug/1 upon overnight standing. Because both the solder and the copper piping must be exposed, however, galvanic corrosion is usually a more serious problem with new plumbing. Many studies have shown that the age of the lead solder is among the most important variables affecting solubility. For example, Sharrett et al. (1982a) -- studying Seattle, a city with few lead pipes found that the age of the house (a proxy measure for the age of the solder and other plumbing materials) was the dominant factor for predicting the concentration of lead in the tap water. In homes that were newer than five years old, with copper pipes, the median lead concentration for standing water was 31 ug/1 versus 4.4 ug/1 in older homes. In homes built within the previous 18 months, the median lead level was 74 ug/1. New solder will leach lead even in relatively non-corrosive water -- whether naturally less corrosive or treated (e.g., Nielsen, II-8

1976; Herrera et al., 1981) and will continue to leach signifi- cant amounts for up to five years (Lovell et al., 1978, Sharrett et al., 1982a; Oliphant, 1982 and 1983; Murrell, 1984; Lassovszky,

1985; Neff and Schock, 1985; etc.). Murrell (1985) found that new solder could leach sufficient lead to contaminate standing water to a degree hundreds of times higher than the current MCL; this has been confirmed by data collected by the Philadelphia water utility (1985) and elsewhere. Oliphant (1983) has also shown that no matter how small the area of exposed solder, pro­ vided the contact time is long enough, the lead levels will always exceed 100 ug/1 if the volume of the sample is small.

With new (exposed) solder, the duration of contact need not be long to raise lead levels significantly. Britton and Richards (1981) and Lyon and Lenihan (1977) have shown that, with particu­ larly corrosive drinking water, lead levels in standing water in systems with copper plumbing joined with lead solder could rise

above 100 ug/1 within 40 minutes of contact. Oliphant (1~83) has presented evidence that those conditions can produce lead levels one to two orders-of-magnitude higher than expected from equilib­

rium solubility calculations. Two other factors will affect the rate of lead leaching from

lead-soldered joints: the surface area of the lead/tin solder at the joints (which can often relate to the quality of the plumbing

and jointing work) and the number of joints per length of pipe

(e.g., Nielsen, 1975 and 1976; Lyon and Lenihan, 1977; Walker and

Oliphant, 1982b; Oliphant, 1983; Birden et al., 1985; Lassovszky, II-9

1985; Murrell, 1985; Treweek et al., 1985). It has also been

suggested that with pipes running vertically, the lead solder can drip down the pipe, resulting in more exposed solder than in hori­ zontal pipes (Arthur Perler, u.s. EPA, private communication). Lead solder in brass kitchen faucets can result in particu­ larly high concentrations of lead (Samuels and Meranger, 1984).

Only one paper that we are aware of (Thompson and Sosnin,

1985) suggests that lead solder does not contribute significant

amounts of lead to water. In that article, although the measure-

-ments are presented as first-draw·standing levels after a 12-hour

exposure period, the data shown are really for flushed (running)

samples ( termed 'steady-·state values'). The article was based on

a report (Battelle, 1982) that contains the actual data on lead

levels. In that report, first-flush samples averaged above 50 ug/1 (Figures 22, and 23, and Table 50 of the Battelle report).

This is acknowledged in the narrative on page 33 of the report,

but discounted because the lead content decreases with time.

Therefore, the results of the Battelle study are consistent with

the rest of the literature. The easy plumbosolvency of lead solder has been known for many years. As a result, the Netherlands banned lead solder in

1977, Germany banned the use of lead solder about the same time·,

and several states and localities in the United States have banned it within the past few years. The 1986 Amendments to the

Safe Drinking Water Act prohibit the use of materials containing

lead in public water systems, a prohibition that is enforceable

by the States after June 1988. II-10

II.A.1.c. Lead Pipes Several characteristics of lead piping also influence lead levels in drinking water (PococK, 1980; Britton and Richards,. 1981; etc.). The length of the lead pipe, in both the home and the supply lines, can have a positive association with lead levels Kuch and Wagner, 1983; Department of the Environment, 1977; Pocock, 1980; Karalekas, 1984) as can the position of the pipe -- it appears that, even for the same length of pipe, water composition, etc., lead piping contained wholly or partly within the house (as opposed to lead service connections outside the house) correlates with higher first-draw lead levels. The ratio of the surface area of lead exposed to the water volume contained is also important (Ainsworth et al., 1977). The age of the dwelling or the pipe (Ainsworth et al., 1977) and the percentage of lead piping in both the service mains and within the residence are significant in determining lead levels, as well. New lead pipes appear to leach higher levels of lead initially, with the rate decreasing within a few days or weeks (Ainsworth et al., 1977).

II.A.l.d. Other Potential Sources of Lead Lyon and Lenihan (1977) and others have found that the flux used for soldering is an excellent electrolyte and can contribute significant amounts of lead to drinking water.

Some lead can also leach from copper pipes themselves (Herrera et al., 1981). Specifications for copper pipes usually limit only copper and phosphorus, and copper used for non-drinking water applications is permitted to contain some lead. Copper pipe manufacturers have indicated that copper tubing used for II-11 water can be made from recycled copper products, thus permitting the introduction of lead impurities (Herrera et al., 1981).

Although not common, lead impurities can also occur in galvanized pipe (e.g., Nielsen, 1975).

Lead is also used in the production of brass and bronze.

Brass is a copper-zinc alloy, which can contain up to 12 percent lead, and bronze is a copper-tin alloy, which can contain up to

15 percent lead (U.S. EPA, 1982b). While both are relatively corrosion resistant and are not generally recognized as a source of lead, several studies document lead leaching from brass or bronze fixtures (Nielsen, 1975; Samuels and Meranger, 1984; ·

Neff 1984; Neff and Schock, 1985; Neff et al., 1987).

Lead can also contaminate potable water when used in pipe jointing compounds and through its use for goosenecks, valves, and gaskets in water treatment plants or distribution mains.

At least one study (Herrera et al., 1982) found that lead can leach from tin-antimony solder "presumably com[ing] from impurities in the solder."

Early tests of plastic pipes showed that lead contamination resulting from stabilizers used with polyvinyl chloride (PVC) pipes could be high (studies summarized in National Academy of Sciences,

1982; volume 4, pages 64ff). However, since then, the National

Sanitation Foundation has developed a standardized £esting pro­ cedure for plastic pipes.

Additional analyses of the leaching of lead from these and other sources is needed (AWWA-DVGW Report, 1985). II-12

II.A.l.e. Other Factors Relating to Lead Contamination Levels

The most important other variable affecting lead levels in drinking water is the length of contact time between the water and the plumbing materials. High lead levels are found in water from faucets that are seldom used or in the first drawn samples in the morning after the water has sat overnight. Usually, flushing the faucet will significantly decrease the lead levels in the water. This is true for all waters, corrosive and less corrosive.

With very corrosive waters or new solder, the length of contact

time need not be great -- as little as 40 minutes to an hour can

produce lead levels above 100 ug/1 under certain conditions

(e.g., Lyon and Lenihan, 1977; Britton and Richards, 1980; Kuch

and Wagner, 1983; Oliphant, 1983). The number of occupants of the dwelling is inversely propor­

tional to lead levels, probably because fewer occupants means the

water will, on average, remain in the pipes longer (Department

of the Environment, 1977; Pocock, 1980). Other factors that can affect lead levels in tap water are

pipe length and diameter, and water velocity (Bailey and Russell,

1981; Kuch and Wagner, 1983). No matter what metal the pipe is

made of, the diameter of the pipes is inversely proportional to

lead contamination levels because of the greater proportion of

water in contact with the lead-containing surface (solder, flux

or pipe) in pipes with smaller diameters (Crank, 1975). The

length of the metal pipe can correlate directly with the potential for high lead concentrations although not consistently (Sharrett II-13 et al., 1982a; Kuch and Wagner, 1983). Lead levels can also increase with the turbulence and velocity of the water, and with irregular flow patterns.

II.A.l.f. Summary

Because lead levels in tap water are a complicated function of the specific qualities of the water, the particular materials comprising the distribution and household plumbing, and the length of time the water is in contact with the plumbing materials as well as other factors, there is a high degree of within-house and between-house variability in water lead levels (Sartor et al., 1981; Bailey and Russell, 1981). With even mildly aggressive

(corrosive) water, however, any amount of lead anywhere in the distribution system or household will contribute some lead to the drinking water. Overall, there are four major risk factors: the the degree of corrosivity, the length of time in the pipe, the total amount of lead in the plumbing materials, and the newness of the plumbing.

In general, lead levels in first-draw water can be several times higher than in running water (e.g., Battelle, 1982). With aggressive waters and new solder, however, first-draw samples can easily be an order-of-magnitude or more greater than running levels (cf. data in Karalekas et al., 1975; Oliphant, 1983;

Maessen et al., 1985; Murrell, 1985; etc.). Lead concentrations in fu-11y flushed samples typical of distribution system water, II-14 even under corrosive conditions and with new solder, are generally below 50 ug/1 and are usually below 20 ug/1.*

II.A.2. Prevalence of Lead Materials in Distribution Systems

Lead has been used for plumbing materials since at least

Roman times. It was considered a convenient and suitable conveyance of water and was used extensively for water pipes during the nineteenth and early twentieth centuries.

The danger of lead contamination of drinking water was not unknown, however. In 1845, the Report of the Commissioners to

Examine the Sources from Which a Supply of Pure Water May Be

Obtained for the City of Boston concluded, "Considering the deadly nature of lead poison, and the fact that so many natural waters dissolve this metal, it is certainly [in] the cause of safety to avoid, as far as possible, the use of lead pipe for carrying water which is to be used for drinking." Lead pipes were outlawed in severai' German states in the second half of the nineteenth century because of concern over health (cited in AWWA­ DVGW Report, 1985; p. 223). And a warning of potential danger from lead pipes was given to the New England Water Works Associa­ tion in 1900 (Forbes), showing that the risk was known there, as well.

* This pattern of lead contamination also holds in Canada. Data on lead contamination in raw, treated and distributed water from 70 municipalities across Canada show levels averaging 1 ug/1 (Meranger et al., 1979). On the other hand, a study of lead levels in waters that had sat overnight in home plumbing showed 20 percent of the samples exceeding 50 ug/1, with a mean level of 43 ug/1 (Maessen et al., 1985). II-15

Despite these and other warnings, in 1924 Donaldson reported that half of the 539 cities he surveyed in the early 1920s used lead or lead-lined service pipes. He found the greatest usage of lead service lines in New England, New York, the Midwest, Texas,

Oklahoma and Montana.

Because lead is a relatively durable material, many of the original lead lines are still in service (Patterson and O'Brien,

1979).

More recently, Chin and Karalekas (1984) surveyed 153 pubkic water systems in 41 states, Puerto Rico and the District of Columbia, to ascertain the prevalence of lead materials in distribution systems. Their survey targeted large systems, with

91 of the 153 systems surveyed serving populations over 100,000 people. (Nationally, only 0.5 percent of community water systems serve over 100,000 people.) They found that almost three-quarters of the systems had used lead service lines or connections (most of the remaining qu~rter did not know if they had any lead or lead-lined services), and one city (Chicago) still installed lead service lines.* In addition, almost two-thirds of the systems had lead goosenecks in their plumbing (another 10 percent didn't know if they had any) and about half of the systems reported the use of solder or flux containing lead in the distribution system.**

* The installation of new lead pipes is now prohibited in Chicago.

** The utilities may be referring to use in home plumbing or in service lines. II-16

The use of lead/tin solder is ubiquitous in U.S. residential plumbing at present. Indeed, until 1986, most local plumbing and construction codes recommended or even required the use of copper pipe joined by lead/tin solder. The 1986 Amendments to the Safe

Drinking Water Act banned the future use of materials containing lead in public water systems or in residences connected to public systems. The ban became effective immediately (June 19, 1986), but States have up to two years to enforce this provision.

II.B. Data on the Occurrence of Lead in Drinking Water

Under the provisions of the Safe Drinking Water Act, EPA must ensure that public drinking water supplies are free of contamination and that they comply with primary drinking water regulations; this authority includes setting monitoring require­ ments to assess compliance. Sections 1401(1)(D) and 1445(A)(l).

According to EPA regulations, monitoring for inorganic compounds, including lead, must be conducted once per year for systems whose source is surface water and once every three years for water supplies using ground water. 40 CFR §141.23. MCLs are defined as

"the maximum permissible level of a c.ontaminant in water which is delivered to the free flowing outlet of the ultimate user of a public water system. 11 40 CFR §141.1 (a). "Free flowing" has been generally understood by the States and the water utilities to mean a "fully flushed" sample. In addition, the procedures for laboratories certified for reporting purposes under EPA's Laboratory Certification Program, administered by the Office of Drinking Water, specify that the sample be taken after running the water for two II-17 or three minutes (U.S. EPA, 1982c). However, because lead is a corrosion by-product, these procedures decrease the likelihood of detecting lead contamination. Therefore, the data collected for compliance with the Safe Drinking Water Act, as administered currently by EPA, does not adequately represent exposure to lead in U.S. drinking water.

Several studies have investigated the quality of drinking water in the United States, including lead levels (e.g., the

National Inorganics and Radionuclides Survey, the National

Organics Monitoring Survey). But these surveys have also not addressed the phenomenon of most lead contamination of drinking water -- as a corrosion by-product. These surveys have sampled lead levels in fully-flushed water typical of distribution water.

Again, this sampling procedure minimizes the likelihood of detec­ ting the contamination of tap water by lead. Several other studies of national water quality have been conducted over the past two decades (e.g., the 1969 and 1978 Community Water System

Surveys, the Rural Water Survey, the First National Health and

Nutrition Examination Survey), but the results from those surveys have not been used by EPA in setting national standards. (See, for instance, U.S. EPA, 1985; or 40 CFR Part 141, page 46969.)

The estimate of occurrence, therefore, is based upon data collected and analyzed for EPA's Office of Drinking Water in

1979-81. These data portray partly-flushed (30 seconds) kitchen tap samples collected by the Culligan Water-Softening Company;*

* The use of company names and the presentation of related data does not constitute endorsement of their services. II-18

James Patterson of the Illinois Institute of Technology analyzed the data. EPA believes that the Culligan tap samples are more representative of consumed water than are the fully-flushed samples taken in compliance with EPA's monitoring regulations.

In addition, these samples of partly-flushed, random daytime water are more appropriately used with the water-lead-to-blood-lead equation accepted by EPA than would be data on fully flushed water

or even a value integrating average consumption patterns. After the presentation of the Culligan data, this section

also includes a discussion of potential biases in the data, the

consistency of the findings with field and laboratory results,

and an alternative analysis of the potential contamination of drinking water by lead (to confirm the magnitude of the estimates). Finally, because homes with newly-installed plumbing contain­ ing lead solder or flux have a higher risk of elevated lead levels

than homes with older plumbing, this section also discusses

exposure to lead likely to occur in new housing.

II.B.1. Patterson Study/Culligan Data on Tap Water In 1979-1980, EPA's Office of Drinking Water funded a study

by James Patterson using data on residential water quality (Patterson, 1981). The study, "Corrosion in Water Distribution

Systems", analyzed 772 municipal water samples collected by

Culligan Water Softening dealerships in 580 cities in 47 states. The samples were collected from May to November 1978 at the consumers' kitchen taps. No samples were collected from households

using home water softeners. II-19

The purpose of this study was to evaluate the relationship between different corrosion indices and water quality variables likely to influence water corrosivity against the observed levels of the corrosion by-products iron, copper, lead and zinc. The

impetus for EPA's commissioning of this analysis was probably the suggestion of the court in a lawsuit brought by the Environ­ mental Defense Fund against the Agency. EDF v. Castle, 578 F.2d

337, 349~50 (D.C. Cir. 1978). In addition to containing data on the levels of these metals, the analysis contained information on calcium, magnesium, sodium, pH, alkalinity, chloride, conductivity, sulfate, and silica levels in each sample. Calculations of hardness and corrosivity indices (including Langelier, Ryznar,

Aggressive, Driving Force, and Larsons, as well as Dye's Momentary

Excess) are also presented. water samples were collected by Culligan Dealership repre­ sentatives throughout the United States, a£ter the kitchen tap had been flushed for 30 seconds at a moderate flow rate, according to a standardized sampling procedure. All samples were collected

in virgin plastic polypropylene bottles with plastic screw tops. The metals analyses were conducted by the Illinois Institute of Technology and the other analyses were done by the Culligan laboratory in Northbrook, Illinois. Most samples were collected between 10:00 a.m. and 8:00 p.m. For 1-ead determinations, lithium nitrate was added to the acidified sample (although not until the samples were transferred

to glass bottles) and the flameless graphite furnace atomic II-20 absorption procedure was employed. The standard addition technique was used for all lead determinations, to avoid certain common interference problems. Standard Methods {1971) were employed for

the analysis of all samples, except for the atomic absorption procedures, where Perkin-Elmer {1977) procedures were used. Analytical tests were conducted to evaluate potential testing

and analytical biases, both upward and downward. To test for

upward bias {contamination that would increase apparent lead levels), nitric acid was stored in the plastic sample bottles for

two weeks to draw out any lead in the plastic bottle itself,

which could otherwise contaminate the sample. The liquid was

then tested for lead presence, and there was none. Blank samples

(i.e., bottles filled with lead-free distilied water) were also used to check for upward bias. After a period of standing {two

weeks), the distilled water was tested for lead contamination; it

remained lead-free. These efforts showed that contamination of

the samples from the plastic bottle itself was unlikely. On the other hand, because no nitric acid was used to pre­

serve the samples until the water samples were transferred to

glass vials, there was a possibility that some lead from the sample would adsorb to the plastic bottle; this would bias the

results downward {i.e., lower the measurable lead level). Spiked

samples (i.e., solutions with a known amount of lead) were used

to test for this. The results indicated that, on average, 3 ug/1

of lead adsorbed onto the plastic bottle; therefore, a slight

downward bias was present in the results. II-21

Calibration and other analytical controls were also employed at several points in each batch as part of the laboratories' ongoing quality assurance and quality control efforts. (Full documentation is available from each lab.)

These data indicate that 16 percent of partly flushed water samples could exceed an MCL of 20 ug/1 at the kitchen tap, and that 3 percent exceed the current MCL of 50 ug/1. Fifty-two percent of the samples contained 9 ug/1 o.r less of lead, the maximum was 10,000 ug/1. The occurrence of high lead levels was geographically widespread. Samples with lead levels greater than 20 ug/1 were taken in more than half of the states in the country, and in every one of EPA's ten Regions. The distribution of lead levels in these samples is presented in Table II-1.

II.B.l.a. Consistency with Other Data

The literature on contamination levels due to lead leaching from plumbing materials shows great variations relating to the specific conditions being observed.

The Culligan data, with 16 percent of the samples exceeding

20 ug/1, have a lower incidence rate of high lead concentrations than is commonly portrayed in the literature on lead leaching rates and the potential for lead contamination in tap water. This is reasonable because that literature, in general, focuses upon high risk (i.e., very corrosive) waters. II-22

TABLE II-1. Distribution of Culligan Data (Patterson, 1981)

Measured Lead Concentrations (ug/1)

< 10 11-19 20-49 > 50

Percent of samples 60 24 13 3 II-23

The Culligan samples, on the other hand, do not portray particularly corrosive waters. In these samples, the median pH is 7.2 (mean, 6.8), median alkalinity is 106 mg/1 as caco3 (mean, 144 mg/1), and median hardness is 145 mg/1 as CaC03 (mean, 203 mg/1). These are considered to be relatively "non-corrosive" waters. Comparing the hardness of the water in thes_e tap samples, for instance, with the U.S. Geological Survey estimates of the extent of soft water in the United States (Durfor and Becker, 1964a and 1964b) or the data in Millette et al. (1980) ,* these samples represent water that is much harder than the average in the country. The average Langelier Saturation Index** for the Culligan samples is -0.4, which is fairly stable.t It is logical that less corrosive waters would contain a lower incidence of corrosion by-products than do studies of more corrosive water.tt

* Both of these studies are discussed more extensively below in Section B.1.C. ahd in Chapter v. ** One index used to estimate a water's potential corrosivity, discussed more fully in Chapter v. t A "stable" water.is one where a film of CaC03 should be exact­ ly at equilibrium, i.e., it neither dissolves nor deposits and grows. tt While it is hard to generalize about these studies, typical results under corrosive conditions show from perhaps 50-100 percent of the samples exceeding 50 ug/1 (e.g., Lyon and Lenihan, 1977; Department of the Environment, 1977; Britton and Richards, 1980; Oliphant, 1983) to.15-50 percent exceeding 50 ug/1 (e.g., Karalekas et al., 1977; Karalekas et al., 1978; Seattle water Metals survey, 1978; Craun and McCabe,. 1975) • By comparison, the Culligan estimate (16 percent >20 ug/1) is low. II-24

The Culligan data portray higher levels of lead contamination, however, than do eitheL EPA's data on compliance with the Safe

Drinking Water Act or most of the studies that have investigated the quality of distributed drinking water in the United States, including lead levels. These surveys and EPA's monitoring require­ ments have measured lead levels in water more typical of the distribution system and most of the results indicate low levels of lead contamination.* Lead contamination, however, occurs pr~arily in tap water and. in water that has been in contact

(standing) with pipes for some length of time. Therefore, the somewhat higher lead levels in the partially flushed, random daytime tap samples that make up the Culligan samples confirm the findings from experimental and field studies of lead leaching rates and patterns. Perhaps the clearest indication of the consistency of the

Culligan samples with other data on lead contamination of drinking water is a comparison with the preliminary results of EPA's "Lead Solder Aging Study", presented in u.s. EPA (1987). The lead levels in the partially flushed Culligan samples are similar to

those in housing over two years old with median pH (i.e., 7.0-7.4) or in waters with high pH (i.e.,> 8.0). The data from the Lead

Solder Study are presented in Tables II-3 and II-4, below.

* For comparison, the National Inorganics and Radionuclides Survey, recently conducted by EPA, shows only about 1.5 per­ cent of ground-water-supplied public systems have lead levels over 20 ug/1 in fully flushed water. II-25

The Culligan data serve as the basis for this analysis because, of the available data, EPA believes that they are more representative of the water consumed in this country than are the

fully-flushed samples taken in compliance with EPAvs monitoring regulations. People drink the water from the taps in their homes after it has been sitting for unknown periods of time. While some people may run their water before using it for cooking or drinking, probably many people (especially children) do not.

These data portray neither an upper bound for exposure (which would result from analyses of first-flush samples) nor a lower bound (which would come from an analysis of fully-flushed distri­ bution water); they more closely portray the bulk of the water

that is consumed (e.g., Bailey and Russell, 1981). Finally, when controlling for other environmental sources of lead exposure, studies of the relationship between blood lead levels and water lead have also found a better fit with lead levels in standing water than with fully-flushed samples (e.g., Worth et al., 1981; Bailey and Russell, 1981: Pocock et al., 1983) .*

II.B.l.b. Potential Biases in the Data

In determining whether it is reasonable to generalize from a subset, it is important to determine any potential biases in the data and, if found, to assess the likely effect upon the results.

Some potential biases (for example, potential analytical and

* This issue is presented more fully in the section below on uncertainties, within the discussion of the relationship between blood lead levels and levels of lead in drinking water. II-26 testing biases) as well as th_e consistency of the findings with the literature on water quality and lead leaching rates were dis­ cussed in previous sections. Others, for instance, the inclusion of new housing within the data, are discussed below. In this section, we address the issues of potential selection bias, geographic representation, and the implications of the relative hardness of the samples. Because these samples were collected by a water treatment company, a self-selection bias is possible; i.e., it is conceiv­ able that they represent water that is 'dirtier' than average.

However, Culligan is not a general water treatment company and does not test for contamination by inorganic (e.g., lead) or organic (e.g., pesticides) substances. EPA and the Illinois

Institute of Technology arranged for and conducted the analyses of metal contamination in these samples. Homeowners who had general water problems or who had reason to suspect that their water was contaminated would be unlikely to call Culligan for water testing or treatment. If they did call and outline such a problem, the Culligan representative would have either suggested that they contact another lab in the area or informed them that Culligan would charge extra to arrange for such analyses. There­

fore, it is unlikely that there would be selection bias resulting

in higher lead contamination levels. On the other hand, a different selection bias is likely and

did, in fact, occur. Not surprisingly, because the samples were

taken by a water softening company, overall hardness results from II-27

the Culligan/Patterson data were higher than other data on water hardness in the country.* For comparison, only about 10 percent of the Culligan samples contained soft water (i.e.,< 60 mg/1

as CaC03) while in the U.S. Geologi~al survey (Durfor and Becker, 1964a and b) ,** about one-third of the country had soft water. National patterns portrayed in the U.S. Geological Survey held in

this survey: the Northeast and Southeast had the softest water, with the Midwest averaging almost three times harder water. The

Northwest also had relatively soft water.

The highest lead ·levels in this data set were in the Midwest, which also had the hardest water. The combination of high hardness

and high lead levels is somewhat surprising; usually lead (as a

corrosion by-product) occurs more frequently in softer water.

Schock (1980 and 1981) and Schock and Gardels (1983) have shown,

however, that at a pH of 8-8.5, soft water. (i.e., 30 mg/1 as caco3 ) may be less corrosive to both lead and copper than can be hard water (i.e.,> 150 mg/1). These results confirm Patterson's findings (1981) and the consensus of .the technical literature:

Many inter-related factors affect a water's corrosivity and no single corrosivity index adequately measures the actual corrosion

potential of a specific water; therefore, no single index is a good predictor of corrosion by-products, including lead.

* The range in hardness in the Culligan data is extreme: from 2 · to 975 mg/1 as CaC03, with median 145 and mean 203. Generally, levels over 240 mg/1 are quite uncommon. ** The USGS data is discussed more fully in Chapter v. II-28

But while no single measure of water quality is a good

predigtor of the actual corrosion potential of a specific water,

the weight of the literature does show that softer water tends to ., be more corrosive than harder water. Therefore, this selection

bias may introduce some downward bias to the estimates; i.e.,

the concentrations of lead in the Culligan data may be somewhat

low relative to the actual levels in the country as a whole.

Because the people who call Culligan can afford to pay for

those services, another potential self-selection bias is that

the population represented by these data is wealthier {and possibly more sensitive to health and environmental issuses) and contains

a greater proportion of single-family and owner-occupied housing

than is typical in the u.s. population. It is unclear what effect this could have on the estimates. A brief discussion of possible contamination patterns in multi-family housing, particu­

larly high-rise buildings, is included in the section on uncer­

tainties, below.

A question arises as to whether the samples are geographi­

cally representative of the United States. The percentage of

samples collected from each state closely reflects the popula­

tion distribution across the United States, with a few exceptions.

As a proportion of percentage-of-samples to percentage-of-u.s.­ population, the most significant differences are the states of

Alabama, Arkansas, North Carolina and Washington, where the dif­

ference varies by about an order of magnitude. For the state of

Wyoming, the difference is a factor of 5. These states are II-29 relatively small and combined, contain only about 7 percent of the u.s. population. This could be expected because the smaller the size of the observation, the larger the relative effect. The anticipated effect on the estimate is low. For an impact upon the final estimates, the most significant differences are the states of California, Colorado and Illinois, where the absolute difference between the percentage of samples and percentage of population differs by more than 5 percent of the total. On a regional basis, however, the distribution of samples closely paralleled the population distribution. The most significant exception here is the Midwest; the percentage of samples (42 percent) was much larger than the population (26 percent, 1980 census). Table II-2 presents the distribution of samples by state, and by region of the country. We conducted chi-square tests* of the distributions to determine how closely related they were. The results were inconclusive, however, because whether the results were significant depended upon how finely subset the data were, that is, how many divisions (and, therefore, how many degrees of freedom) for the data were used. The possibilities included 50 subsets (i.e., by state, with 49 degrees of freedom), 10 (i.e., by EPA Region, with 9 degrees of freedom), 5 (i.e., using Patterson's groupings, with

4 degrees of freedom), or simply for the nation as a whole.

* The chi-square statistic, represented by the Greek letter x raised to the 2nd power, is a measure of how much, propor­ tionally, the frequencies in the observations differ from the frequencies you would "expect" if there were absolutely no relationship between the variables (Matlack, 1980). II-30

TABLE II-2. Municipal water Sam.ptes and· Population Percentages from Culligan/Patterson Study (Patterson, 1981)

State % Population* % Samples·

Northeast States

CT 1.4 0.63 MA 2.5 2.4 ME 0.4 0 NH 0.4 0.38 NJ 3.3 5.2 NY 7.8 5.3 PA 5. 3_ 3.9 RI 0.4 a.so VT 0.2 0.25 Total 22% 19% southeast States

DE 0.3 0.25 FL 4.3 2.9 GA 2.4 1.8 MD 1.9 1.3 NC 2.6 0.25 SC 1.4 0.63 VA 2.4 2.6 WV 0.9 2.4 Total 16% 13%

Midwest States IA 1.3 3.7 IN 2.4 6.2 IL 5.1 12.4 KS 1.0 1.9 MI 4.1 2.0 MN 1.8 4.2 MO 2.2 2.3 ND 0.3 0.38 NE 0.7 1.3 OH 4.8 4.6 SD 0.3 0.76 WI 2.1. 2.0 Total 26% 42%

* u.s. Bureau of Census, 1980. In the Patterson study, 1978 Census data were used. II-31

TABLE II-2. (Continuecl)

State % Population* % Samples

Southcentral States

AL 1.7 0.25 AR 1.0 0 KY 1.6 1.0 LA 1.9 1.3 MS 1.1 0.63 OK 1.3 0.8R TN 2.0 1.8 TX 6.3 5.2

Total 17% 11%

Western states

AZ 1.2 2.0 CA 10.s 2.3 co 1.3 6.3 ID 0.4 0.51 MT (J. 3 0.88 NM 0.6 0.88 NV 0.4 0.51 OR 1.2 1.3 UT 0.6 0.25 WA 1.8 0.13 WY 0.2 1.0

Total 19% 16%

* Bureau of Census, 1980 II-32

To check whether the slight skewing of number-of-samples

compared to population-by-state affected the estimates, we weighted

the results by state population and compared that analysis with the results of the national analysis (which assumed that the data are geographically representative of the nation). The resu·lts

differed by less than one-half of one percent. We concluded that the slight variation in geographic distribution does not signifi­

cantly affect the estimates.

II.B.l.c. Alternative Analysis of Potential Exposure to Lead in Drinking Water Several bodies of literature are available for an alternative

analysis of potential exposure to confirm the magnitude of the results from the Culligan data. These include studies of the

extent of highly corrosive water in the United States and experi­ mental and field analyses of lead contamination. Two major studies have focused on assessing the extent of highly corrosive water in the United States.* They are the U.S.

Geological Surveys (USGS) conducted in the early 1960s (published in Durfor and Becker, 1964a and 1964b) and the First Health and Nutrition Examination Survey (HANES I), conducted by the National

Center for Health Statistics in 1974 and 1975.** The results of

* This information is presented more fully in Chapter V, section B.l.

** The data from the Midwest Research Institute (1979) and from Millette et al. (1980) on the extent of corrosive water in the country contain profiles that are quite similar to the USGS and the HANES I. They are not included in this discussion, however, because this analysis addresses estimates of the extent of specifically soft water, not water that is corrosive for other reasons. Chapter V discusses these two studies as well. II-33 these studies can be combined with analyses of lead leaching rates and the variables affecting plumbosolvency for comparison with the findings from the Culligan data. The USGS data showed that 17 states had very soft water

(under 60 mg/1 as CaC03); using 1980 Census data, the combined populations of those states is 67.7 million people. The results of the HANES I (as published in Greathouse and Osborne, 1980) show a very similar picture: about one-third of the country has very soft water (under 60 mg/1 as caco3 ). Given a total current (1985) national population of a little over 240 million, about

80 million people receive very soft water. ,

Numerous studies of plumbosolvency conducted in the United

States and in Great Britain have shown that soft, acidic waters are most corrosive and have the highest lead contamination levels (e.g., Craun and McCabeu 1975; Nielsen, 1976; Hoyt et al., 1979;

Patterson and O'Brien, 1979,; U.S. EPA, 1982b; Sheiham and Jackson, 1981; Worth et al., 1981). There is some discussion as to whether pH or carbonate content is the most important variable, and many studies show that the relationship between pH or carbonate and lead levels may be neither simple or linear. In general, the lower either value is, the more vulnerable the water to high lead contamination.* Both laboratory and field experiments demonstrate this.

For instance, data from Karalekas et al. (1977) on two cities

* Because neither the relationship between pH nor carbonate and lead is linear, this is not strictly true at all pH or carbonate levels. Kuch and Wagner (1983) and Schock and Gardels (1983) among others have developed multi-dimensional models of lead solubility. II-34

(Bridgeport, Connecticut, and New Bedfor~, Massachusetts) with similar pH levels (7.1 and 7.3, respectively) but different hardness and alkalinity (48 and 18 mg/1 vs. 12 and 24 mg/1, respectively) indicated that even with moderate pH the softer water could contain six to eight times higher lead contamination levels. Schock and Gardels (1983), investigating water with high pH(> 8.6) but low carbonate content, present average lead concen­ trations of 67-134 ug/1 in first draw water. Other results typical of this literature are the field data presented in the Seattle Water Metals Survey (1978) and the laboratory data in Sheiham and Jackson (1981). In the former, highly corrosive waters produced mean lead levels of almost 30 ug/1 (from lead solder alone; there were no lead pipes) and in the latter, mean lead levels were over 100 ug/1 with a mixture of lead and non-lead pipes (some old, some new) and lead solder. In sum, most studies show that first flush samples from soft, corrosive water (i.e., hardness under 60 mg/1 as CaC03) often result in lead contamination levels exceeding SO ug/1* even in housing that is not new. (New housing is at particular risk of high lead levels. This is discussed in the next section.) Not everyone who receives very soft water, however, is at equal risk of high lead exposure. Some water systems with very corrosive waters (Boston and Seattle, for instance) treat their water to reduce its aggressiveness; their risk is lower. A small proportion of water utilities have lead in source water, as well

* Because EPA's standard for lead is SO ug/1, most studies have used that as the cut-off. II-35

as a corrosion by-product; their risk is higher. Some localities have few lead pipes and connections (and hence, have a smaller

likelihood of high lead contamination), while others have many.

Some localities have begun already to ban the use of materials

containing lead in public water supplies and in residences

connected to them. In addition, plastic pipes are beginning to

be used in some residential plumbing, even for bringing in potable

water,* replacing the metal pipes that are more likely to leach

lead. Finally, at least one city (Akron, Ohio) has instituted an

active program to replace its lead service connections. To account for these variables and for the idiosyncracies of

specific waters, we assumed conservatively that with very soft

waters (i.e., hardness under 60 mg/1 as caco3 ), half of first­ flush samples could exceed 20 ug/1, and that 10 percent of them

could exceed 50 ug/1.

Combining information, then, on 1) the extent of very corrosive

water in the country (about one-third of the population receives very soft water) and 2) studies of potential lead contamination

with very soft water (almost all have levels > 20 ug/1 in first­

flush samples and many also had levels > 20 ug/1 in random daytime

samples) but 3) mitigated to include some factors likely to decrease

exposure* (e.g., some cities have already begun corrosion control

treatment) yields the following calculation:

* Plastic pipes are used more commonly for waste water than for intake. II-36

1/3 of country x 50% occurrence of Pb>20 ug/1 = 17% of population exposed to water >20 ug/1

Therefore, this alternative estimate of potential exposure to lead contamination of drinking water in non-new housing yields results that are quite close to those of the Culligan data ,analyzed by Patterson (1981), where 16 percent of the samples were greater than 20 ug/1. The results are most sensitive to the estimate of lead levels> 20 ug/1. (For comparison, assuming 25% occurrence yields a national estimate of 8% and assuming 75% occurrence yields a national estimate of 25% of the U.S. population exposed to water> 20 ug/1.) There is additional anecdotal evidence that also supports the patterns and extent of lead contamination presented here. A CBS-affiliated television station in Cleveland, Ohio (Channel

8) conducted a small survey of lead levels in first-flush tap water in early 1987. They found 13 percent of the samples exceeded

20 ug/1, with a home occupied for only three :r;nonths having lead levels of~ 100 ug/1. Second, a water utility in Colorado (Little Thompson Water District) conducted a limited survey of lead contamination within customers' residences. This survey was conducted in conjunction w1th the eolorado Department of Heal~h and was accomplished concurrently with sampling programs in the cities of Denver, Colorado Springs, and Fort Collins (Colorado).

The results from Little Thompson showed one-third of the residences sampled exceeded 20 ug/1 -- results they indicated "were comparable

* Some factors, for instance, many lead pipes or lead contamination of distribution water, can increase exposure estimates, also. II-37 to the results of the aforementioned cities." The highest reading

(> 400 ~g/1) was in a not-yet-occupied house. Third, concern about potential lead contamination of drinking water in Washington,

D.C. in late 1986 and early 1987 resulted in about 1,000 water samples being taken there. Newspaper reports (Washington Post) and numerous press releiases from the D. C. Department of Public

Health reported that between 13 and 25 percent of the water samples exceeded 20 ug/1, and that the results indicated "that as many as 56,000 houses may have problems with lead contamination," resulting from the common use of long lead service connections in many parts of the city. In addition, data collected by KYW-TV in Philadelphia (an NBS-affiliate) in February 1987, by EPA's

Region IX (San Francisco) in Spring 1987, by the Nassau County (New York) Department of· Health in 1987, and jointly by the New Jersey

Department of Environmental Protection and the U.S. Geological

Survey (presented by J. L. Barringer at the American Geophysical Union Spring 1987 meeting) show widespread lead contamination following predictable patterns: contamination levels are high with new plumbing or corrosive waters, contamination is minimal with non-corrosive waters and older plumbing. Finally, data available from WaterTest Corporation,* a private water testing laboratory in Manchester, New Hampshire, on over 2,500 samples taken in January-March, 1987, show that average

first-flush samples exceeded 20 ug/1 in 15 states plus the District of Columbia. This shows that the occurrence of lead levels exceeding 20 ug/1 is widespread in this country but it was not

* The use of company names and the presentation of related data does not constitute endorsement of their services. II-38 possible to determine the age of the housing from which the samples came. This alternative analysis of potential lead contamination of tap water was based upon assumptions about leaching rates in soft water(< 60 mg/1 as CaC03), but this is not the only factor that makes water corrosive. Many other parameters, including pH, alkalinity, temperature, etc., contribute to corrosivity. People who receive water that is only moderately soft (i.e., hardness between 60 and 90 mg/1 as CaC03) as well as those whose 45ter is not soft but has a low pH or has other risk factors associated with aggressiveness (e.g., high levels of chlorine, dissolved oxygen, chlorides, sulfates, etc.) also run a somewhat elevated risk of exposure to high lead levels in drinking water.

However, the data on these circumstances are too sparce to use in estimating populations at risk of high lead levels.

II.B.2. Lead Contamination in New Housing As was discussed briefly in section A above, many studies have shown that newly-installed lead solder (or pipes) can leach high amounts of lead in a short amount of time. Indeed, in studies such as Sharrett et al. (1982a), the age ·of the house (a proxy measure of the age of the plumbing) was the variable most closely related to the lead concentration in the house water. Table II-3 presents some field data on lead levels in new housing. As can be seen, the highest lead contamination levels occur with the new­ est solder (i.e., during the first 24 months following installation those levels decline and generally are not elevated beyond five years (cf also, Sharrett et al., 1982a; Lassovszky, 1984; etc.). II-39 ,

TABLE II-3 Lead Contamination Levels in Tap Water by Age of Plumbing (Field Studies} Mean Pb % of Age of Level Samples Study Housing (ug/1) >2D ug/1 Conditions/Notes

Sharrett­ <18 months 74 NG (Median standing levels et al. <5 years 31 NG (No lead pipes (1982a) >5 years 4.4 NG (

Nassau unoccupied 2,690 NG (Average, first flush - County (1985) <2 years 540 NG ( 2-10 years 60 NG ( >10 years 10 NG (

Philadelphia <2 years 90 NG Flushed (1985) 5000 NG First-flush

>2 years 60 NG Flushed 500 NG First-flush

>4.5 years <25 NG NG

EPA (1987) <2 years NG 93% First draw (pH <6.4) preliminary NG 51% Flushed, 2 minutes (pH <6.4) results <2 years NG 83% First draw (pH 7.0-7.4) NG 5% Flushed, 2 minutes (pH 7.0-7.4}

<2 years NG 72% First draw (pH >8.0) NG 0% Flushed, 2 minutes (pH >8.0}

2-5 years NG 84% First draw (pH <6.4) NG 19% Flushed, 2 minutes (pH <6.4)

2-5 years NG 28% First draw (pH 7.0-7.4) NG 7% Flushed, 2 minutes (pH 7 .0-7 •.4)

2-5 years NG 18% First draw (pH >8.0) NG 4% Flushed, 2 minutes (pH >8.0)

>5 years NG 51% First draw (pH <6.4) NG 4% Flushed, 2 minutes (pH <6.4}

>5 years NG 14% First draw (pH 7.0-7.4) NG 0% Flushed, 2 minutes (pH 7.0-7.4)

>5 years NG 13% First draw (pH >8.0} NG 3% Flushed, 2 minutes (pH >8.0}

KEY: Pb= Lead NG = Not given II-40

Oliphant (1982 and 1983), studying the galvanic action between lead solder and copper pipes, concluded that with any amount of new (exposed) solder.alone (i.e., without lead pipes) and sufficient time, lead levels will always eventually exceed

100 ug/1 if the volume of the sample is small. The parameters of the water (including pH and total carbonate) are irrelevant to the prediction. His analysis also showed that this galvanic action can produce lead contamination levels 100-1,000 times higher than equilibrium models would predict for the water itself.

Over time, however, protective passivation films usually build up in the plumbing and such galvanic couples can stabilize eventually at between 10 and 50 percent of the highest (initial) leaching rates (Oliphant, 1983) or even lower (Lyon and Lenihan, 1977).

With copper pipes and new solder, flushed water samples can exceed the current MCL (Philadelphia, 1985; Nassau, 1985; Kuch and Wagner, 1983), although this is not common; the length of new plumbing is probably a significant factor here.

While new housing containing lead solder clearly rep~esents a significant risk of extremely high lead level~ in drinking water, the possibility existed that this risk was included in the samples collected by Culligan. We evaluated this in two ways. First, we compared the incidence rate in the Culligan data (16 percent of the samples exceeded 20 ug/1) against the rates pre­ sented in studies of new plumbing (about 100 percent> 20 ug/1).

The literature on the leaching potential of new plumbing shows much higher contamination levels than is evident in the Culligan kitchen tap samples. Therefore, the data indicate that it is unlikely that these samples included new housing. Second, we II-41 contacted Culligan dealerships in eight different areas, including both areas of recent rapid growth and areas with relatively stable growth patterns. In each office, we asked the representative with the longest sales record to describe the average Cul.ligan customer on a public water supply and to estimate how many customers on public water supplies lived in new homes. The representatives contacted had served many thousands of customers and all described their customers as generally calling Culligan with iong-standing problems; for residences connected to community water systems, new problems and problems in new homes were perceived as going first to the local public water utility. Inhabitants of new housing {i.e., built within the past two years), therefore, represent a separate group at risk of receiving water that exceeds 20 ug/1. Because this population is not represented in the samples of partially flushed kitchen tap water, they must be added to the estimate of exposure.

II.C. Estimated Exposure to Lead in U.S. Tap Water The estimate of exposure to lead levels in U.S. drinking water~ 20 ug/1 has several components: 1) the general risk of high lead levels due to the corrosivity of all water and the con­ tact time between tap water and any materials containing lead, 2) the specific risk to inhabitants of new homes {built within the past two years), and 3) many assumptions, including the generaliz­ ability of the findings, the distribution of the at-risk popula­ tions, the relationship between lead levels in water and human blood-lead levels, compliance with a new regulation, and other issues. II-42

The first part of this section describes some of the many uncertainties and assumptions included in this analysis. The second part presents estimates of potential exposure to tap water containing 20 ug/1 of lead or greater.

II.C.l. Uncertainties and Assumptions in the Analysis It is important to note some of the assumptions and uncer­ tainties that are both inherent and explicit in this analysis.

Assumption: Total compliance. This analysis assumes that, should the MCL be reduced, all community water systems will comply with the new standard by whatever means are necessary for that par- ticular system. In reality, if some systems do not comply, both the costs and the benefits will be overestimated proportionately: the benefit to cost ratio will remain the same because both are functions of the number of people affected. No more than borderline compliance with the new standard is assumed, however. That is, if the MCL is 20 ug/1, we assume that any systems currently exceeding that level will take measures to reduce their water to that level. If some systems act to reduce lead levels further, both the costs and the benefits will be higher; it is unclear what the ratio between the costs and benefits would be for the incremental reduction.*

* The case study of Boston (Jacobson, 1986), summarized in Chapter I and appended to this document, indicates that in particular circumstances, the benefit to cost ratio may be quite high 11:1 in that case. It is unclear, however, whether or how Jacobson's results can be extrapolated to other U.S. water systems and cities and, therefore, to this proposed rule.

It is unlikely, though, that there is any reasonable and practical scenario in which a water system would institute corrosion control measures where the incremental costs exceed the incremental benefits. At a minimum, therefore, these costs and benefits should be assumed to be equal. II-43

Assumption: Treatment can generally reduce lead levels to 20 ug/1. In general, the data on corrosion control treatment show signi­ ficant reductions in lead levels (e.g., Karalekas et al., 1977,

1978, 1983; Herrera et al., 1983). However, efforts have generally focused upon reducing lead contamination to levels below the current MCL of 50 ug/1. There is very little data on reducing lead levels below that. Some preliminary data is available from EPA's "Lead Solder Aging Project." These results, shown as Table II-4, indicate that simply raising pH alone greatly reduces the occurrence of samples exceeding 20 ug/1. Field data from Patterson and O'Brien (1979), Britton and

Richards (1981) and others show that adjusting the alkalinity of the water also significantly affects plumbosolvency. Other studies (e.g., Schock and Gardels, 1983) adjusted the level of dissolved inorganic carbonate (DIC) as opposed to alkalinity, because DIC is independent of pH while alkalinity is not. These studies found that increasing DIC could slow lead leaching rates. Some laboratory, experimental, and theoretical analyses have addressed the effect of simultaneously altering several water parameters (e.g., pH, alkalinity, inorganic carbonate, etc.) on plumbosolvency (e.g., Schock, 1980; Jackson and Sheiham, 1980;

Schock and Gardels, 1983; Sheiham and Jackson, 1981; Kuch and

Wagner, 1983). However, no treatment has yet been shown to be completely successful in preventing all contamination of drinking water by lead. In particular, new (exposed) plumbing containing lead II-44

'mBLE II-4. Percentage of Samples Exceeding 20 ug/1 of Lead at Different pH Levels, by Age of House Percent of samples >20 ug/1 Age of House pH First-flush Fully-flushed (2 min) 0-2 years <6.4 93% 51% 7.0 - 7.4 83% 5% >8.0 7,2% 0%

2-5 years <6.4 84% 19% 7.0 - 7.4 28% 7% )8.0 18% 4%

6+ years <6.4 51% 4% 7.0 - 7.4 14% 0% <8.0 13% 3%

Source: U.s. EPA (1987) , preliminary results fran "Lead Solder Aging Study" II-45 seems to continue to produce relatively high levels of contami­ nation even in relatively non-corrosive waters (e.g., Neilsen, 1976}. The preliminary results from the EPA solder study on pH adjustment, however, indicate that even these levels can be reduced significantly through treatment. For this analysis, we assumed that the best treatment cur­ rently available will be adequate to reduce lead levels to

20 ug/1, except perhaps in certain specific circumstances such as first-flush waters that have been standing for 16 hours or more, or with lead solder that is under two years old.

Assumption: EPA will change its monitoring requirements to better detect corrosion by-products in drinking water. The common inter­ pretation of EPA's current monitoring requirements calls for fully-flushed samples, typical of distribution water.* such prac­ tice will not capture the presence of lead in drinking water, or indeed, the presence of any corrosion by-products, and results in underestimations of contamination and exposure. This analysis assumes that EPA will change its regulations to capture that exposure and that the criteria for compliance with the new MCL will consider the risk of contamination by corrosion by-products. A revision of EPA's monitoring requirements was called for by the court in the decision on a lawsuit brought by the Environmental

Defense Fund against the Agency, as EPA accumulated data on the

* This issue is discussed at the beginning of Section rr.B. II-46 factors likely to affect lead levels in drinking water. EDF v. costle, 578 F. 2d 337, 349-350 (p.c. Cir. 1978). It was also clearly the intent of the National Primary Drinking Water Regula­ tions (40 CFR §141.2(c)) and the Preamble to the Regulations (FR, volume 40, number 248, p. 59575 - December 24, 1975) that corrosion by-products be addressed. These revisions have been expected by the regulated community (e.g., AWWA Committee Report, 1984) and by professionals in the field (e.g., Hoyt et al., 1979~ Patterson and O'Brien, 1979) for several years.

Assumption: Adults consume 2 liters of water (or water-based fluids) per day and children consume 1 liter per day. For consistency with past analyses, this document used the estimates that are commonly held in the published literature and that have served as exposure indexes in past EPA actions.

Assumption: Lead levels are tap specific. Because the level of lead contamination depends largely upon the length of contact time between the water and the plumbing as well as the parameters of the water, the particular materials of the private and public plumbing systems, and the age of the plumbing, lead levels vary from tap to tap within a system and even within a house.

Assumption: The relationship between lead in drinking water and human blood-lead levels. The published literature presents several possible equations relating lead levels in human blood to intake of lead from drinking water. This analysis uses the II-47 formulae presented in the Quantification of Toxicological Effects Section of EPA's Water Criteria Document (1985) for lead, which are taken from the Air Quality Criteria Document (U.S. EPA, 1986):

(for children) PbB* = 0.16** x intake of lead from water

( for adults) PbB* = 0.06** x intake of lead from water.

The coefficient for children is taken from Ryu (1983), and the source of the coefficient for adults is Pocock (1983). Both assumed a linear relationship between lead in drinking water and blood lead levels. Two other general approaches exist. The constant used for children (PbB = 0.16 x Pb in water) is derived from a study of infant blood lead levels (Ryu, 1983) ,. in which the "constant" was really a non-steady state value. This may be an inappropriate value to use or may be, at best, a lower bound estimate.t A better "constant" may be the steady state value from the control group in the Ryu study, which was 0.45. If so, the projections of children's health effects in this analysis may be underestimated by as much as a factor of 3.

Other studies. of the relationship between blood lead levels and water lead concentrations (cf. the discussion and bibliographic citations in the Air Quality Criteria Document, 1986, p. II-106ff.)

* PbB = blood lead level ** These constants have a unit of ug/dl per ug/day. t The authors of the Air Quality Criteria Document and the Water Criteria Document are aware of this problem. II-48 for both children and, more commonly, adults, have found cube root functions, typically with intercepts of 4-7 ug/dl (blood lead, with water lead= 0). While the linear forms cited above and used in this analysis provide results similar to the non-linear forms over a very wide range of values (up to, say, 300 ug/1 of water), the form of the model greatly influences the estimated contributions to blood lead levels from relatively low water-lead concentrations. over the typical range of lead levels in u.s. drinking water (0-100 ug/1), the differences in estimated blood lead levels can be quite large. Indeed, in the range of this analysis (generally, water lead levels of 0-50 ug/1), various cube-root functions yield values that are 4-10 times greater than

the estimates using the linear form presented above. Alternative

assumptions (e.g., those reasonably derived from the results of

Richards and Moore, 1982 or 1984) could indicate that exposure -­

and consequently benefits -- in this analysis may be underestimated,

possibly by several factors. studies investigating the relationship between lead in

drinking water and lead in the blood have found a better fit between blood lead levels and lead levels in standing or first­

flush samples than other measures of lead contamination of water,

for instance, fully flushed samples (Worth et al., 1981; Pocock

et al., 1983; Bailey and Russell, 1981). This seems counter­

intuitive at first because the intake of standing water would be

expected to be much less in total volume consumed than for other

water (e.g., partly or fully flushed water), so the expected II-49 contribution of the relatively high lead levels typical of standing samples should be small. But several studies (Rabinowitz et al.,

1976 and 1980, in particular) have shown that the absorption of lead varies depending on the state of the gastrointestinal system.

Specifically, lead ingested on an empty stomach (e.g., at breakfast or between meals) has a much higher absorption rate than does lead ingested on a full stomach. This could explain the closer correlation between blood lead levels and lead levels in standing or first-flush water.

Where studies have related blood lead levels to several different measures of drinking water lead, the correlation coefficient is larger for the running samples than for the first­ flush samples, generally by a factor of 1.5-2 (e.g., Worth et al.,

19811 Central Directorate on Environmental Pollution Study, 1982). Whatever measure of water lead is used (first-flush, fully-flushed, etc.), the corresponding coefficient of relationship to blood lead must be used. That means, if lead levels in partly flushed water are measured, the coefficient should also be for partly flushed samples.* Howe·\7er, no studies calculated the relationship between blood lead levels and lead levels in partly flushed water. This analysis, therefore, uses the best.available analyses, the Ryu and Pocock studies discussed above, which evaluated first-flush water. The use of the coefficients for first-flush

* No study has yet calculated a value relating blood lead levels to an integrative measure of water consumption patterns, i.e., a measure reflecting actual drinking habits. A new epidemiological study to derive such a coefficient would be necessary in order to use data on actual consumption patterns. II-50 water with occurrence data relating to contamination levels in partly flushed water introduces a downward bias to the estimates.

In general, the studies assessing the form of the blood-lead/ drinking-water-lead relationship, especially the British studies, assume no contribution to the body burden of lead from any other environmental sources besides drinking water. (The major exception is Worth et al., 1981.) For infants, this may be a less significant omission than for toddlers or adults. Curiously, the authors of these studies do not question why there is an intercept of 4-7 ug/dl, even if water lead is zero. However, gasoline lead is an important determinant of human blood-lead levels (cf. Air Quality

Criteria Document for Lead, 1986, p. 11-42ff; Chapter 3 of The

Costs and Benefits of Reducing Lead in Gasoline, 1985; and sources cited there). Indeed, the reduction of gasoline lead levels in the United States in the late 1970s appears to have resulted in a reduction in children's blood-lead levels of almost half during that period (Annest et al., 1983). Lead paint, under certain conditions, can also result in high localized contamination levels. Analyses of the contributions from various media to human blood-lead levels, focusing upon exposures typical in this country, indicate that drinking water lead may account for about

14-55 percent of the total burden of lead. It is most likely that drinking water contributes 15-40 percent of the lead body burden (cf, discussions throughout the

Air Quality Criteria Document, summarized pp. 13-26ff). II-51

This analysis employs the Ryu and Pocock coefficients (dis­ cussed above) relating lead levels in drinking water to blood lead levels; these are combined with the drinking water lead

levels presented in this chapter to calculate the potential effect upon blood lead levels. These changes are projected onto

extrapolations of the data from the Second National Health and

Nutrition Examination survey (NHANES II)* on the distribution of blood lead levels in the country (cf, Air Quality Criteria Document, chapter 11) to predict the health benefits that would result from

a potential reduction in the MCL from 50 ug/1 to 20 ug/1.

Assumption: New solder containing lead (under 24 months) contri­

butes an average of 25 ug/1 of lead to drinking water. Many

field and laboratory studies have found that lead solder alone

when used with copper household plumbing -- could easily produce lead levels in drinking water well above the current MCL, even in

relatively non--corrosive waters. To be conservative, we assumed

* · The NHANES II was a 10,000 person representative sample of the u.s. non-institutionalized population, aged 6 months to 74 years. The survey was conducted by the (U.S.) National Center for Health Statistics (NCHS) over a 4-year period (1976-1980). The data base is available from NCHS and analyses of the lead-related data from it have been published before (e.g., Annest et al., 1982 and 1983; Mahaffey et al., 1982a and 1982b; Pirkle and Annest, 1984).

These extrapolations incorporate the reductions in exposure resulting from the current phasedown in lead in gsoline (the limit is currently 0.l grams of lead per gallon of gasoline). The projections developed in support of that rule-making and presented in The Costs and Benefits of Reducing Lead in Gasoline (U.S. EPA, 1985b) form the basis of the projections developed for this potential rule. II-52 that new solder would on average produce lead levels at half the

MCL, i.e., 25 ug/1, in partly flushed tap water.

Assumption: People drink partly flushed tap water.* Lead levels are highest in first-flush samples, that is, in water that has been sitting for several hours or more (for instance, overnight or all day). But those conditions occur only infrequently (at most, once or twice per day for each faucet), and the sparce data available on actual drinking water use patterns indicate that the bulk of consumed water is partly flushed (e.g., Bailey and Russell, 1981). The likelihood of flushing the water before using it probably follows age and sex patterns, and those most at risk of lead's adverse health effects-~ children -- may be least likely to flush the water.

Therefore, EPA has concluded that people are more likely to consume partly flushed water than fully flushed water.

On the other hand, two particular demographic trends over the past few decades are likely to result in an increase in the amount of lead from drinking water to which people are exposed. First, the number and proportion of women working outside the home has increased from 40 percent in 1970 to 51.9 percent in 1983 (Statistical Abstracts, 1985: Tables 27, 658, 659, and elsewhere}. This means that more homes will have two 'first­ flushes' per day -- one in the morning and the other when the

* Or its equivalent: Some first-flush and some fully flushed water. II-53 parent(s) return from work.* Obviously, this doubles the possi­ bility of exposure to high first-flush lead levels. The second demographic trend that is significant is the decrease in the average number of occupants per housing unit. In 1950, there were, on average, 3.37 people per housing unit, which has decreased to 2.73 in 1983 (Statistical Abstracts, 1985; Table 54). The number of occupants of ct dwelling is inversely proportional to lead levels in the drinking water, probably because fewer occupants mean the water will, on average, remain in the pipes longer.

In addition, as noted above in the discuss ion of the rela­ tionship between blood lead levels and levels of lead in drinking water, blood lead levels correspond best with first-flush or standing levels.

Uncertainty: Factors affecting lead levels in multi-family housing, especially high-rise buildings. The available data on lead contamination at the tap comes primarily from single-family homes. It is unclear how lead concentrations in multi-family housing will vary. In the absence of good data, there are three hypothesized possibilities: lead levels will be higher in high-rises than in single-family homes with similar water, lead levels will be lower, or contamination levels and patterns will be the same. There are plausible arguments to support each thesis.

Lead levels may be higher in high-rise buildings because it is more difficult, and perhaps impossible, to fully flush the

* Exposure to lead in the work place or at school is not included in this analysis. II-54 water in a building that is more than a few stories. Because the water is never really flushed, average residence time in pipes is longer and therefore contamination levels will be greater.

On the other hand, lead levels may be lower because one of the factors contributing to high lead levels in tap water is the relatively close ratio of water volume to surface area of pipe resulting from the narrow pipes typical of home plumbing. While the pipes going to each faucet may be comparable, high-rise buildings have service pipes that are, on average, much wider than the largest pipes in homes. Because of the reduced ratio of water to pipe surface area, the potential for high lead contamina­ tion may be lower. Another factor, number of occupants, also cor­ relates inversely with lead levels. Because multi-family housing has more occupants in total, average residence time for water in pipes may be shorter than in single family homes.

The third possibility is that both arguments above are correct, and that they cancel out. Contamination levels in large apartment buildings would be comparable to levels in single-family homes.

II.C.2. Calculations of Exposure to Lead in Drinking Water Several adjustments to .the available data on lead levels in

tap water and in new housing are necessary to predict the number of people served by community water supplies who are likely to be

exposed to drinking water exceeding an MCL of 20 ug/1. These

include the assumptions discussed above and other data on the II-55 composition of the housing stock in the United States and the use of various plumbing materials. These estimates are for one sample year only, 1988 •. Because the ban. o.n the use of materials containing lead in public water supplies will become enforceable after June 1988, exposure to lead in new housing can be expected to begin to decreasa thereafter. rr.c.2.a. Estimate of Exposure to Lead in Drinking Water to Inhabitants of New Housing

The published literature shows that inhabitants of new housing are at risk of exposure to high levels of lead in drinking water. The rates are highest for the first two years, but they decline and are generally not elevated beyond five years.

There were 1.7 million new housing starts and permits in the

United States during 1983 and 1.8 million in 1984.* Construction data show that housing typically takes six months to a year from permit to potential occupancy, so there are cur!ently about 3.5 million new housing units (i.e.,< 24 months). The Statistical Abstract of the United States (1985) indicates that in 1983, the average household contained 2.73 individuals (Table 58). Multi­ plied together, a total of 9.6 million people currently live in new housing.

However, not all of these people are served by community water supplies. Of the current (1985) u.s. residential population

* Survey of Current Businessv u.s. Department of Commerce - Bureau of Economic Analysis, 1985i Table on New Housing Construction. II-56

(a little over 240 million), 219.2 million are served by community water systems and this analysis only addresses that population.

In addition, the use of plastic plumbing materials has increased recently and voluntary switching to lead-free solder has occurred in many areas; these homes are obviously at decreased risk of exposure to lead from the leaching of new lead/tin solder. Data from the plumbing supply industry show that about 8 percent* of new plumbing is plastic,** so 92 percent of the population has metal pipes. We assumed that virtually all those with metal pipes also have some solder or other fittings containing lead, although this may overestimate exposure somewhat. We assumed that the inhabitants of new housing are distributed proportionately between community and non-community water supplies. Therefore, the number of people served by community water supplies at risk of high lead levels from new solder in new housing is:

219 million 9.6 mil x 240 million x .92 = 8.1 million people.

This estimate, based upon the current population and current building practices, is for one sample year, 1988. The ban on the future use of materials containing lead in public water supplies

* This is the arithmetic average of claims by the Plastic Pipe Institute presented in Mruk (1984) and of the Copper Develop­ ment Association presented in Anderson (1984). ** Plastic pipes are used more commonly for waste water than for intake water. But the use of plastic pipes for both is increasing rapidly. II-57 and in residences connected to them, established by the Safe

Drinking Water Act as amended in 1986, will be enforceable after

June 1988. Exposure to lead in new housing can be expected to decrease thereafter.

II.C.2.b. Estimate of Exposure to Lead in Drinking Water to Inhabitants of Older Housing

The data on partly flushed kitchen tap samples indicate that

16 percent of the drinking water in housing older than two years in this country may exceed an MCL of 20 ug/1. To avoid double counting, the inhabitants of new housing served by community drinking water systems must be subtracted from the total number of people served by community water supplies.

Of the current (1985) U.S. residential population of a little over 240 million, 219.2 million are served by community water systems. Again, we assumed that the peopie who live in new housing (9.6 million) are distributed proportionately between community and non-community water supplies. Therefore,

219 million 9.6 million x 240 million =·8.8 million people.

To calculate the risk to inhabitants of older housing, subtract the number in new housing (8.8 million) from the total served by community water systems (219.2 million); that indicates that

210.4 million people live in older homes. Based upon the Culligan data, 16 percent of them (33.7 million) are at-risk of high leaq levels in partly flushed water at their kitchen taps. II-58 rr.c.2.c. Total Estimated Exposure to Lead in Drinking Water Combining the avaflable data on lead levels in older housing

(33.7 million people exposed to lead levels> 20 ug/1) with the new housing exposure estimates (8.1 million people at risk) indicates that 41.8 million people using public water supplies currently may be exposed to some water that exceeds the proposed

MCL of 20 ug/1; we round this to 42 million. This estimate is for one sample year, 1988. Many uncertain­ ties surround this estimate, indicating that it may be high or low. overall, exposure to lead in drinking water is expected to decrease somewhat after 1988 because of the Congressional ban on

the future use of pipes, solder and flux containing lead in public water systems and in residences connected to them.

On the other hand, this may be a low estimate

0 because it does not include the potential exposure of occupants in housing built within the past 2-5 years;*

0 because we have not included those who, while living in older housing, have recently had major plumbing repairs and so are also at risk of the potentially high lead levels associated with newly-installed solder;

0 because the Culligan data represent water that is harder than average, whereas high lead levels are often found with soft waters;

* The incremental risk to inhabitants of 2-5 year old housing is not included in this analysis because it was not possible to eliminate those people from the base and thus avqid double­ counting. II-59

0 becaus~ the data used are for partially flushed samples, while some people (especially children) may consume water that is closer to first-flush or standing samples;* and

0 because some of the statistical and analytical techniques used lend a downward bias to the results (e.g., the method of sample preservation and the use of a fir~t-flush correlation coefficient with data on lead levels in partly-flushed water).

In addition, we ha,,e not included any data from the estimated

60 million people served wholly or in part by private and non­ community water supplie.s.**

* Blood lead levels are more clos~ly related to lead levels in fir.st f1ush or standing water.

** Some people are served by both public and private supplies. For instance, a person's home may be served by a public water system while the drinking water supply at school or ~ork may be private.

CHAPTER III

BENEFITS OF REDUCING CHILDREN'S EXPOSURE TO LEAD

The scientific literature presents evidence of a variety of physiological effects associated with exposure to lead, ranging from relatively subtle changes in various biochemical measure­ ments at very low levels of exposure, to severe retardation and even death at very high levels of exposure. Although such effects are found in individuals of all ages, particular concern has focused qn children.

Because the body is a complex structure of interdependent

systems and processes, effects upon one component will have cascad­

ing implications throughout the body. This interdependence is well illustrated by multi-organ impacts resulting from the inhibi­ tion of heme synthesis by lead, with consequent reduction in the body heme pool. These effects are depicted graphically in Figure III-1, taken from EPA's most recent Air Quality Criteria Document

for Lead {1986; p. 13-31). A summary of children's health effects

from exposure to lead, taken from the Air Quality Criteria Document and included in the Water Criteria Document for Lead {1985), p. VIII-65, is also included here as Figure III-2. This chapter summarizes the available evidence of the effects of lead on children, and estimates some of the health benefits of

reducing exposure by reducing lead concentrations in drinking water. Section A deals with the pathophysiological effects of lead, while Section B addresses the evidence on neuropsychological effects {primarily reduced cognitive ability), and Section C discusses the fetal effects of lead exposure. Section D presents III-2

Figure III-J. MuJ.ti-Qrgan Impacts of Lead's Effects on the Heme Pool

REDUCTION DF HEIi£ IDDY POOL

ERYTHRDPDIETIC REDUCED ANEIIIA - REDUCED EXACERBATION OF CARDIOVASCULAR EFFECTS HEMOGLDIIN OXVG£N TRANSPORT HYPOXIC EFFECts Of DYSFUNCTION ANO SYNTHESIS TD ALL TISSUES OTHER Sf RESS AGENTS DTHER HYPOXIC EFFECTS

EFFECTS ON NEURONS, AXDNS,AND SCHWANN CELLS

1--N.;.;E;.;;U.;.;.RA.;.;l;.._-I~ REDUCED HEIIDPADTEINS IMPAIRED MVELINATIDN EFFECTS (1.1., CYTDCHRDMESI AND NEAVE CONDUCTION

IIIPAIAED OEVELDl'MENT OF NERVOUS SYSTEM

DISTURBED IMUNO IMPAIRED MINERAL tMPAIAED IONE ANO REGULATORY ROLE Of CALCIUM TISSUE HOMEOSTASIS TOOTH DEVELOPMENT

EN AL-ENDOCRINE REDUCED 1,25-IDHlz - DISTUAIED CALCIUM IMPAIRED CALCIUM ADLE AS SECOND EFFECTS VITAMIN D IIETAIDLISM IIESSENGER

DISTURBED ADLE IN IMPAIRED CALCIUM TUMDAIGENESIS ROLE IN CYCLIC ~DNTADL NUCLEOTIDE METAIOLISM

IMPAIRED DETOXIFICATIOII OF ENVl.,ONMENTAl TOXINS IMPAIRED DETOXIFICATION OF XENOIIOTICS IMPAIRED OETOXIFICATIOI OF DRUGS REDUCED HEME FOR HlPATIC HEME REGULATED EFFECTS TRANSFORMATIONS ELEVATED IRA.IN ALTERED METAICl ISM LEVELS OF TRYPTDPHAN, OF TRYPTOPHAN SEROTONIN, ANO HIAA IMPAIRED METAIOLISM Of ENDOGENOUS AGOJflSTS IMPAIRED DIS!URlfiJ IN00UAMINE HYDROXYLATION NEUROTRANSMITTER OF CDRtlSOL FUNCTION

Multi-organ impact of reductions of heme body pool by lead. Impairment of heme synthesis by lead results in disruption of a wide variety of important physio- logical processes in many organs and tissues. Particularly well documented are erythropoietic, neural, renal-endocrine, and hepatic effects indicated above by solid arrows ( ~>. Plausible further consequences of heme synthesis interference by .lead which remain to be more conclu­ sively established are indicated by dashed arrows(----). III-3

FIGURE III-2. SUMMARY OF LOWEST OBSERVED EFFECT LEVELS FOR KEY LEAD-INDUCED HEAL'IH EFFECTS IN CHILDREN

Lowest Observed Herne Synthesis and 13 Neurological Renal System Gastrointestinal Effect Level (PbB) Hematol~ical Effects Effects Effects Effects 80-100 ug/dl Encephalopathic Chronic nephropathy Colic, other overt signs and syrnptans (aminoaciduria, etc.) gastrointestinal symptans I 70 ug/dl Frank anemia I V 60 ug/dl Peripheral neuropathies

50 ug/dl ? 40 ug/dl Reduced hanoglobin CNS cognitive effects synthesis (IQ deficits, etc.) Elevated coproporphyrin Peripheral nerve dysfunction (slowed NCV's) Increased urinary AIA

30 ug/dl Vitamin D metabolism interference I 15 ug/dl Erythrocyte Altered CNS I protoporphyrin electrophysiological I elevation responses I I I 10 ug/dl AIA-D inhibition I I V V Py-5-N activity ? ? inhibition I V ? Abbreviations: PbB = blood lead concentrations; Py-5-N = pyrimidine-5 '-micleotidase; CNS = central nervous system; NCV = nerve conduction velocit; AIA = arninolevulinic acid pource: Air Quality Criteria Document for Lead (1986) III-4 the methods used to monetize the benefits of reducing children's exposure to lead. Section E discusses some inherent limitations of cost-of-illness studies, upon which most of these monetized health benefits are based, and lists many additional health effects that are not included at all in our analysis. A SU!'.llffiary of the monetized and non-monetized children's health benefits for one sample ,year (1988) is presented in Section F.

Fuller discussions of the adverse health effects associated with lead exposure can be found in other EPA documents: the Air

Quality Criteria for Lead (1986), including the Addendum which is part of Volume 1; the Quantification of Toxicological Effects section of the Drinking Water Criteria Document on Lead (1985); and The Costs and Benefits of Reducing Lead in Gasoline (1985).

This document rests heavily upon the analysis and methodologies in The Costs and Benefits of Reducing Lead in

Gasoline. Two sections (III.A.4 on Stature Effects and III.C. on Fetal Effects)" have been expanded from the earlier analysis; the other sections have been condensed. This reflects the inclusion of new materials and is not an indication of relative importance. In addition, this document includes an alternative method for valuing one aspect of cognitive damage: potential decrement in IQ, valued as a function of the potential decrease in future earnings. This chapter also contains a discussion of the limitations of cost-of-illness studies, both in general and of the specific studies which serve as the basis of this analysis, that did not appear in the earlier cost benefit analysis. III-5

The estimates of health benefits associated with this proposed rule rely upon data on the distribution of lead levels in children and adults collected as part of the Second National

Health and Nutrition Evaluation Survey (NHANES II). The NHANES II was a 10,000 person representative sample of the U.S. non-institu­ tionalized population, aged 6 months to 74 years. The survey was conducted by the (U.S.) National Center for Health Statistics (NCHS) over a four-year period (1976-1980). The data base is available from NCHS and analyses of the lead-related data from it have been published before (e.g., Annest et al~, 1982 and 19837

Mahaffey et al., 1982a and 1982b7 Pirkle and Annest, 1984). This survey provides careful blood, biochemical, nutritional and many other biological, social, and demographic measures representative of the u.s. population. The fact that other sources of lead, especially gasoline, would slowly decline even without new EPA drinking water standards created a slight complication in projecting blood lead levels for sample year 1988. Because gasoline lead levels fall over time as unleaded gasoline replaces leaded, the difference in blood lead levels resulting from this rule will change over time. The estimates in this report account for both reductions in some other sources of lead and changes in the demographic profile of the u.s. population. This model served EPA also in its analytical efforts supporting the most recent phasedown in the amount of lead permitted in leaded gasoline; it is discussed more fully in

The Costs and Benefits· of Reducing Lead in Gasoline (US-EPA,

1985b). III-6

III.A. Pathophysiological Effects

Elevated blood-lead levels have long been associated with neurotoxicological effects and many other pathological phenomena:

an article on lead's neurotoxicity was published as early as 1839,

on kidney damage in 1862, and on impaired reproductive function

in 1860. From an historical perspective, lead exposure levels

considered acceptable for either occupationally-exposed persons or

the general population have been revised downward steadily as more sophisticated biomedical techniques have shown formerly­ unrecognized biological effects, and as concern has increased

regarding the medical and social significance of such effects.

In the most recent downward revision of maximum safe levels (late

1984 - early 1985), the Centers for Disease Control (CDC) lowered

its definition of lead toxicity to 25 ug/dl blood lead and 35

ug/dl of erythrocyte protoporphyrin (EP). The present literature shows biological effects as low as 10 ug/dl (for heme biosynthesis)

or even 6 ug/dl (for fetal effects and for IQ effects in some

populations); indeed, some effects (e.g., elevated ALA levels,

hearing decrements, or stature effects) have exhibited no threshold

so far.

There is no convincing evidence that lead has any beneficial

biological effect in humans (Expert Committee on Trace Metal

Essentiality, 1983).

The finding of biological effects at the lowest observed blood-lead levels (4-6 ug/dl) potentially has important implica­

tions for public health, because such levels are common in the III-7

U.S. population. As Table III-1 shows, between 1976 and 1980 over three-quarters of children under the age of 18 had blood lead ievels in excess of 10 ug/dl, and 15 percent exceeded 20 ug/dl. The rates among blacks and among preschool children were even higher.

Lead's diverse biological effects on humans and animals are seen at the subcellular level of organellar structures and processes, and at the overall level of general functioning that encompasses all of the bodily systems operating in a coordinated, interdependent way. The biological basis of lead toxicity is its ability, as a metallic cation, to bind to bio-molecular substances crucial to normal physiological functions, thereby interfering with these functions. Specific biochemical mechanisms include lead's competition with essential metals for binding sites, inhibition of enzyme activity, and inhibition or alteration of essential ion transport. The effects of lead on certain subcellu­ lar organelles, which result in biochemical derangements common to and affecting many tissues and organ systems (e.g., the disrup­ tion of heme synthesis processes), are the origin of many of the diverse types of lead-based functional disruptions of organ systems.

Lead is associated with a continuum of pathophysiological effects across a broad range of exposures. In addition to the high level effects mentioned above, there is evidence that low blood-lead levels result in:

1. Inhibition of pyrimidine-5'-nucleotidase (Py-5-N) and delta-aminolevulinic acid dehydrase (ALA-D) activity, which appears to begin at 10 ug/dl of III-8

TABLE III-1. Blood Lead Levels of Children in the United States 1976-80 (percent in each cell~ rows sum to 100 percent) 10-19 20-29 30-39 40-69 <10 ug/dl ug/dl ug/dl ug/dl ug/dl

All Races all ages 22.1 62.9 13.0 1.6 0.3 6 months-5 years 12.2 63.3 20.5 3.5 0.4 6-17 years 27. 6 64.8 7.1 0.5 o.o

White all ages 23.3 62.8 12.2 1.5 0.3 6 months-5 years 14.5 67.5 16.1 1.8 0.2 6-17 years 30.4 63.4 5.8 0.4 o.o

Black all ages 4.0 59.6 31.0 4.1 1.3 6 months-5 years 2.7 48.8 35.1 11.1 2.4 6-17 years 8.0 69.9 21.1 1.0 o.o III-9

blood lead or below (Angle et al., 1982). Hernberg and Nikkanen (1970) found 50 percent of ALA-D inhibited at about 16 ug/dl.

Inhibition of erythrocyte ALA-D appears to occur at virtually all blood lead levels mea~ured so far, an9 any threshold remains to be determined (cf, summary of literature in Air Quality Criteria Document, 1986; pp. 12-13 to 12-51).

2. Elevated levels of EP zinc protoporphyrin (ZPP) in red blood cells at about 15 ug/dl. This probably indicates a general interference in heme synthesis throughout the body, including interference in the functioning of mitochondria (Piomelli et al., 1977). Changes in heme metabolism have been reported peri­ natally at blood lead levels of 8-10 ug/dl {Lauwerys et al., 1978). Some studies that accounted for iron status show that children with low iron stores are more sensitive to lead in terms of heme biosynthesis interference (e.g., Mahaffey and Annest, 1986).

3. Changes in the electrophysiological functioning of the nervous system. This includes changes in slow­ wave electroencephalogram (EEG) patterns and increased latencies in brainstem auditory evoked potentials (Otto et al., 1981, 1982, 1984) which begin to occur at about 15 ug/dl. The changes in slow-wave EEG patterns appear to persist over a two-year period. Also, the relative amplitude of synchronized EEG between left and right lobe shows effects starting at about 15 ug/dl (-Benignus et al., 1981). Finally, there is a significant negative correlation between blood lead and nerve conduction velocity in children whose blood lead levels range from 15 ug/dl to about 90 ug/dl (Landrigan et al., 1976).

4. Inhibition of globin synthesis, which begins to appear at approximately 20 ug/dl (White and Harvey, 1972; Dresner et al., 1982). 5. Increased levels of aminolevulinic acid (ALA) in blood and soft tissue, which appear to occur at about 15 ug/dl and may occur at lower levels (Meredith et al., 1978). Several studies indicated that·increases of ALA in the brain interfered with the gamma-aminobutyric acid (GABA} neurotransmitter system in several ways (Criteria Document, 1986; p. 12-145 ff).

6. Inhibition of vitamin D pathways, which has been detected at the lowest observed blood-lead levels (Rosen et al., 1980a, 1980b; Mahaffey et al., 1982). Further, as blood lead levels increase, the inhibition becomes increasingly severe. III-10

7. An inverse relation between maternal and fetal blood-lead levels and gestational age, birth weight, and early post-natal development (both physical and mental) down to 10 ug/dl and possibly below (Bellinger et al., 1984; McMichael et al., 1986). Investigations of post-natal growth and development also present evidence of a negative association with blood-lead levels at the lowest observed blood-lead level (Schwartz et al., 1986).

8. Finally, recent studies of IQ effects in poor black children (Schroeder, 1985; Schroeder and Hawk, 1986) show IQ effects over the range of 6 to 47 ug/dl, without an evident threshold (cf also Air Quality Criteria Document, 1986; p. 12-92 ff, 12-157, and elsewhere). Another recent article (Schwartz and Otto, 1987) shows hearing effects throughout the range of measured blood-lead levels. These data cite the lowest observed effect levels to date, and do not necessarily represent affirmative findings of thresholds below which exposures can be considered safe. The specific effects listed above as occurring at blood lead levels below 25 ug/dl indicate (a) a generalized lead impact on erythrocytic pyrimidine metabolism, (b) a generalized lead-induced inhibition of heme synthesis, (c) lead-induced interference with vitamin D metabolism, and (d) lead-induced perturbations in central and peripheral nervous system functioning.

As lead exposure increases, the effects become more pro­ nounced and broaden to additional biochemical and physiological mechanisms in various tissues, causing more severe disruptions of the normal functioning of many organ systems. At very high lead exposures, the neurotoxicity and other pathophysiological changes can result in death or, in some cases of non-fatal lead poisoning, long-lasting sequelae such as mental retardation and severe kidney disease. III-11

This chapter discusses the known pathophysiological effects of lead that occur in children, particul~rly the neurotoxic and fetal effects, and the expected change in the number of children at potential risk of those effects under EPA's proposed drinking water regulation.

III.A.I. Effects of Lead on Pyrimidine Metabolism

The best-known effect of lead on erythrocytic pyrimidine metabolism is the.pronounced inhibition of Py-5-N activity, an enzyme that controls the degradation and removal of nucleic acid

from the maturing red blood cell (reticulocyte). As noted earlier,

the disruption of this function by lead has been noted at exposure

levels beginning at 10 ug/dl. At blood lead levels of 30-40 ug/dl, this disturbance is sufficient to materially contribute to red blood cell destruction and, possibly, decreased hemoglobin production contributing to anemia (World Health Organization, 1977; .National Academy of Sciences, 1972; Lin-Fu, 1973; Betts et

al., 1973). The mechanism of this inhibition may have a wide­

spread impact on all organs and tissues.

III.A.2. Effects on Heme Synthesis and Related Hematological Processes

These effects, are described more fully in the Air Quality Criteria for Lead (EPA, 1986), are only summarized here. III-12

III.A.2.a. Mitochondrial Effects

The mitochondrion is an organelle within the cell and outside

the nucleus that produces energy for the cell through cellular

respiration and is rich in fats, proteins and enzymes. It is the

critical target organelle for lead toxicity in a variety of cell

and tissue types, followed probably by cellular and intracellular membranes. The scientific literature shows evidence of both

structural injury to the mitochondrion (e.g., Goyer and Rhyne,

1973; Fowler, 1978; Fowler et al., 1980; Bull, 1980; Pounds et

al., 1982a and 1982b) and impairment of basic cellular energetics

and other mitochondrial functions (e.g., Bull et al., 1975; Bull,

1977, 1980; Holtzman et al., 1981; Silbergeld et al., 1980). These and other studies also provide evidence of uptake of lead

by mitochondria in vivo and in ·vitr·o.

III.A.2.b. Heme Synthesis Effects

The best-documented effects of lead are upon heme biosynthesis.

Heme, in addition ~o being a constituent of hemoglobin, is an

obligatory constituent for diverse hemoproteins in all tissues, both neural and non-neural. Hemoproteins have important roles in

generalized functions, such as cellular energetics, as well as in more specific functions such as oxygen transport and detoxification

of toxic foreign substances (e.g., the mixed-function oxidase

system in the liver). Statistically significant effects are

detectable at 10-15 ug/dl.

The interference of lead with heme synthesis in liver mito­

chondria appears to result in the reduced capacity of the liver

to break down tryptophan, which, in turn, appears to increase III-13 levels of tryptophan and serotonin in the brain (Litman and Correia, 1983). This may account for,some of the neurotoxic effects of lead.

The elevation of aminolevulinic acid (ALA) levels is another indication of lead's interference in heme synthesis and mitochon­ drial functioning. Because increased ALA is associated with significant inhibition of certain kinds of neurotransmission, such elevations can have serious neurotoxic implications. Thus, in addition to its direct molecular neurotoxicity, lead may adversely affect the brain at low exposure levels by altering heme synthesis (e.g., Silbergeld et al., 1982). There appears to be no threshold concentration for ALA at the neuronal synapse below which presyn­ aptic inhibition of GABA release ceases.

Since ALA passes the blood brain barrier and is taken up by brain tissue, it is likely that elevated ALA levels in the blood correspond to elevated ALA levels in the brain (Moore and Meredith, 1976). Furthermore, lead in the brain is likely to enhance brain ALA concentrations because neurons are rich·in mito­ chondria, the subcellular site of ALA production. As mentioned earlier, blood ALA elevations are detectable at 18 ug/dl of blood lead (Meredith et al., 1978).

III.A.3. Lead's Interference with Vitamin D Metabolism and Associated Physiological Processes Another potentially serious consequence of lead exposure is the impairment of the biosynthesis of the active vitamin D metabolite, l,25-(OH)2 vitamin D, detectable at blood lead levels of 10-15 ug/dl. Further, an inverse dose-response rela­ tionship has been reported between blood lead and l,25-(OH)2 III-14 vitamin D throughout the range of measured blood lead values up to 120 ug/dl (Criteria Document, p. 12-37 ff.1 Rosen et al., 1980a, 1980b1 Mahaffey et al., 1982b). Interference with vitamin D production disrupts calcium and phosphorous homeostasis, par­ tially resulting in the reduced absorption of these elements from the gastro-intestinal tract. This may alter the availability of these elements for physiological processes crucial to the normal functioning of many tissues, cell membranes, and organ systems.

The reduced uptake and utilization of calcium has two compounding consequences. First, it interferes with calcium­ dependent processes that are essential to the functioning of nerve cells, endocrine cells, muscle cells (including those in

the heart and other components of the cardiovascular system), bone cells, and most other types of cells. The second concern

is possible increased lead absorption resulting from decreased calcium availability. The latter can be expected to further exacerbate the inhibition of vitamin D metabolism and reduced

calcium availability (Sorrell et al., 19771 Mahaffey et al., 1986), resulting in even greater lead absorption and greater vulnerability

to increasingly more severe lead-induced health effects (Rosen et al., 1980b1 Barton et al., 1978). These effects are especially

dangerous for young (preschool age) children who are developing

rapidly. These children, even in the absence of lead, require a

relatively high intake of calcium to support the formation of the skeletal system, as well as several other calcium-dependent physiological processes important in young children. III-15

Even moderate levels of lead exposure in children are associated with vitamin D disturbances that parallel certain meta­ bolic disorders and other disease states, as well as severe kidney dysfunction {Criteria Document, 1986; p. 12-37}. At blood lead levels of 33-55 ug/dl, l,25-{OH}2 vitamin Dis reduced to levels comparable to those observed in children who have severe renal insufficiency with the loss of about two-thirds of their normal kidney function {Rosen et al., 1980a; Rosen and Chesney, 1983;

Chesney et al., 1983}. Analogous vitamin D hormone depressions are found in vitamin D-dependent rickets {type I}, oxalosis, hormone-deficient hypoparathyroidism, and aluminum intoxication in children undergoing total parenteral nutrition.

Lead-induced interference with 1,25-{OH}2 vitamin D biosyn­ thesis affects a wide range of physiological processes. The vitamin D-endocrine system is responsible in large part for the maintenance of extra- and intra-cellular calcium homeostasis {Rasmussen and Waisman, 1983; Wong, 1983; Shlossman et al., 1982;

Rosen and Chesney, 1983}. Thus, modulation in cellular calcium metabolism induced by lead at relatively low concentrations may potentially disturb multiple functions of different tissues that depend upon calcium as a second messenger {Criteria Document, p.

12-40). It also appears that 1,25-{OH}2 vitamin D participates directly in bone turnover by orchestrating the population of cells within the bone {Criteria Document, p. 12-41}. An immuno­ regulatory role for the vitami~ D hormone is evident through the widespread existence of 1,25-{OH)2 vitamin D3 receptor sites in III-16 imrnunoregulatory cells, such as monocytes and activated lymphocytes

(Provvedini et al., 1983; Bhalla et al., 1983).

The negative correlation between blood lead and serum l,25-(OH)2 vitamin D, the hormonally active form of vitamin D, appears to be another example of lead's disruption of mitochondrial activity at low concentrations. While serum levels of 1,25-(OH)2 vitamin D decreased continuously as blood lead levels increased from the lowest measured level (12 ug/dl), this was not true for its precursor, 25-(OH) vitamin D. In fact, in lead-intoxicated children after chelation therapy, 1,25~(OH)2 ·vitamin D levels were restored, but the precursor levels remained unchanged (Rosen et al., 1980a, 1980b; Mahaffey et al., 1982). This indicates that lead inhibits renal 1-hydroxylase, the kidney enzyme that converts the precursor to the active form of vitamin D. These observations in children are supported by lead effects on vitamin

D metabolism in vivo and in vitro (Smith et al., 1981; Edelstein et al., 1984). Renal 1-hydroxylase is a mitochondrial enzyme system, which is mediated by the hemoprotein, cytochrome P-450.

This suggests that the damage to the mitochondrial systems detected at 15 ug/dl and below has uncompensated consequences.

If cytochrome P-450 is being inhibited at the low levels of blood lead that the reduced renal 1-hydroxylase activity suggests, it is possible that other physiological functions related to cytochrome P-450 are also disrupted. For example, reduced P-450 content has been correlated with impaired activity of the liver detoxifying enzymes, aniline hydroxylase and aminopyrine demethy­ lase, which help to detoxify various medications and xenobiotics III-17 and modulate the metabolism of steroid hormones (Goldberg et al.,

1978; Saenger et al., 1984).

While cytochrome P-450 inhibition has been found in animals, and in humans at higher lead levels, this has not yet been examined in children at blood lead levels< 25 ug/dl. But the disruption of vitamin D biosynthetic pathways at these levels is suggestive of an effect. The reduction in heme caused by lead exposure probably underlies the effects seen in vitamin D metabolism. This would explain the similarity in the effect of lead on both erythrocyte protoporphyrin accumulation and decreases in levels of serum l,25-(OH)2D. It would also indicate a cascade of biological effects among many organ and physiological systems of the body

(depicted graphically in Figure III-1). Together, the inter­ relationships of calcium and lead metabolism, lead's effects on l,25-(OH)2D, and the apparent disruption of the cytochrome P-450 enzyme system provide a single molecular and mechanistic basis for Aub et al.'s observation in 1926 that "lead foliows the calcium stream."

III.A.4. Stature Effects Small stature has been identified with lead poisoning for many years (e.g., Nye, 1929), and is a plausible outcome, given the known biotoxic interaction of lead with calcium messengers, heme-dependent enzymes, and neuroendocrine function. Several new studies provide evidence of a much stronger assoctl.ation III-18 between exposure to lead and subsequent growth and development than was previously thought. These studies of stature effects were not included in The Costs and Benefits of Reducing Lead in

Gasoline. The Addendum to the Criteria Document (1986), appended to Volume 1, contains a discussion of the deleterious effect of lead upon various aspects of development and growth, even at the relatively low exposure levels encountered by the general population, i.e., 15 ug/dl and below (p. A-31 to A-56).

III.A.4.a. Effects of Lead on Fetal Growth

Many studies have investigated the effect of intrauterine

lead exposure on gestational age, fetal growth and fetal physical development.* The Air Quality Criteria Document (1986: pp. 12- 192 to 12-220) and the Addendum to the Criteria Document (1986: pp. A-31 to A-56) contain a full review of these studies.

Several studies examined the relationship between maternal or fetal blood-lead levels and gestational age. Moore et al.

(1982), for instance, conducted a cross-sectional study of 236 mother-infant pairs in Glascow, Scotland. Blood lead levels showed a significant negative relation to gestational age, for both maternal and cord lead measures. The blood lead levels were within the normal range ~nd higher, with geometric mean blood-lead

* Many of these studies also show a relationship between blood lead levels (both maternal and fetal) and negative pregnancy outcomes, including early membrane rupture, miscarriages and spontaneous abortions, potential minor congenital anomalies in live births, etc. These and other adverse effects upon the fetus are discussed in Section III.C. of this report. III-19 levels of 14 ug/dl for the mothers and 12 ug/dl for the infants.

As an indication of the findings, the mean blood-lead levels for the 11 cases of premature birth (gestational age under 38 weeks) were among the highest and averaged about 21 ug/dl for mothers and 17 ug/dl for infants. In this study, first-flush household water lead levels were positively associated with both maternal and fetal blood-lead levels.

In another recent study of gestational age (McMichael et al.,

1986), following 774 pregnancies to completion (live birth, spontaneous abortion, or still birth), women with blood lead levels> 14 ug/dl were over 4 times more likely to deliver pre-term than women with blood lead levels of< 8 ug/dl. Excluding cases of still births, the relative risk increased to almost 9.

Other studies have looked at the relationship between prenatal exposure to lead and birth weight or size. Nordstrom et al. (1979b), examining records of female employees of a Swedish smelter, found decreased birth weights related to: 1) employment of the mother at the smelter during pregnancy, 2) distance the mother lived from the smelter; and 3) proximity of the mother's job to the actual smelting process. In~ related study (Nordstrom et al., 1978a), similar results were found for infants born to mothers merely living near the smelter.

A more recent paper by Bryce-Smith (1986) found both birth weight and head circumference related to placental lead levels in a cohort of 100 normal infants. Another study (Bellinger et al., 1984), studying mental development in middle class children III-20 up to 2 years old, found a more subtle exposure-related trend in the percentage of small-for-gestational-age infants. Dietrich et al. (1986), presenting interim results, also found that pre­ natal lead levels were associated with both reduced gestational age and reduced birth weight, which in turn were both signifi­ cantly associated with reduced neurobehavioral performance at three months.

Some other studie~ (e.g., Clark, 19771 McMichael et al.,

1986*) did not find birth weight statistically significantly related to blood lead levels, however. Nonetheless, "the evidence as a whole from these studies indicates that gestational age appears to be reduced as prenatal lead exposure increases, even at blood lead levels below 15 ug/dl" (Addendum to the Criteria

Document, 19861 p. A-45).

Other recent studies of lead's adverse effect upon physical development have assessed neurobehavioral aspects of child development. These studies are described and evaluated in Section

Bon lead's neurological effects. Lead's adverse impact on gestational age, however, has cascading effects upon subsequent mental development in infants.

III.A.4.b. Effects of Lead on Post-Natal Growth

The first article on lead's effect on stature (Nye, 1929) observed the incidence of "runting", eye squint and drop foot as physical characteristics of overtly lead-poisoned children.

* The Addendum to the Criteria Document (19861 p. A-43f) suggests that the findings of McMichael et al. are "not entirely clear with regard to birth weight. The proportion of pregnancies resulting in low-birthweight singleton infants [in the high blood-lead group] ••• was more than twice that for the [low blood-lead group]." III-21

Since then, however, surprisingly few studies have investigated this effect, until quite recently.

In the 1970s, three studies (Mooty et al., 1975; Johnson and

Tenuta, 1979; Routh et al., 1979) investigated possible stunting of physical growth as an end point of lead exposure. In the first study, the children in the high-lead group (blood leads of

50-80 ug/dl) were shorter and weighed less than those in the low-lead group (blood leads, 10-25 ug/dl). But the high-lead group was also slightly younger (average age 33 months vs. average age 34 months in the low-lead group) and not racially matched, so it is difficult to determine clearly the r~lative contribution of lead to the difference in stature. Johnson and Tenuta studied the growth and diets of 43 low-income children and also found a relative decrease in height with an increase in blood lead level. But they did not report the specific racial composition and mean

·ages of the subjects, nor .did they assess the relative contribution of differences in calcium intake or the incidence of pica or other factors. Routh and co-workers found the incidence of microencephaly (defined as head circumference at or below the third percentile for the child's age on standard growth cha~ts) was markedly greater among children with blood lead levels> 30 ug/dl than in those with blood lead levels below 29 ug/dl.

Again, however; it is not possible to distinguish the relative contribution of lead from other (racial, dietary, etc.) factors that may have affected these children's growth. Despite their individual weaknesses, the three studies together are suggestive of an effect. III-22

Much stronger evidence for the retardation of growth and decreased stature associated with exposure to lead has emerged more recently from animal toxicology studies and the evaluation of large epidemiologic data sets.

About 65 papers on animal experimental studies have been published in the last 10 years that investigated the retardation of growth following low-level exposure during intrauterine life, early post-natal life or both. These studies found decreased body weight at blood lead levels of 18-48 ug/dl with no change in food consumption (e.g., Grant et al., 1980). Deficits in the rate of neurobehavioral development and indications of specific organic or functional alterations were observed at blood lead levels as low as 20 ug/dl (e.g., Fowler et al., 1980). As summarized in the Addendum to the Criteria Document (1986; p. A-

51), "it seems very clear [from these animal studies] that low­ level chronic lead during pre- and early post-natal development does indeed result in retarded growth even in the absence of overt signs of lead poisoning."

Finally, Schwartz et al. (1986) analyzed results from the

Second National Health Assessment and Nutritional Evaluation Survey

(NHANES II) to investigate the relationship between blood lead levels and physical development, controlling for other contri­ buting factors, including age, race, sex, several measures of nutritional status, family income, degree of urbanization and many other variables enabling them to account for general health, and environmental and nutritional factors that might not be adequately controlled for by the nutrient and blood measurements. III-23

To assure that blood lead was not found to be significantly associated with growth and development because of the correlation between blood lead and nutri tio-nal status, a stepwise. regress._ion procedure employing potential confounding variables was used. To address the NHANES II survey design, the computer program SURREGR was used. Schwartz's results show that blood lead levels are a statistically significant predictor of children's height (p < 0.0001), weight (p < 0.001) and chest circumference (p < 0.026), after controlling for age (in months), race, sex and nutritional covariates. Figures III-3 and III-4 illustrate the relationship of stature (height and weight) to blood lead, after controlling for all of the other covariates. The threshold regressions (using segmented regression models) indicate that there is no identified threshold for the relationship down to the lowest observed blood lead of 4 ug/dl. The relations~·lip is consistent through the normal range (5-35 ug/dl). III-l4

FIGURE III-3 Relationship of B.Lood Lead Level to Weight in Children Aged O to 7

ADJUSTED WEIGHT IK1• 17 2

170 • 1&8 •

1&6

164

162

160

15.I

ADJUSTED BLOOD LEAD (ug/dl) Adjusted weight and adjusted blood lead levels for children aged 7 years and younger in Second National Health and Nutrition Examination Survey. Both weight and blood lead level have been adjusted by regression for effects of age, race, sex, and all other variables significant at .05 level. Each point is mean weight and mean blood lead level of approximately 70 consecutive observations, ordered by blood lead levels. Regression line reflects slope of coefficient obtained from multiple regression analysis of all 1,967 obser­ vations with no missing data.

$Qul:;-ce~ Schwa;rtz et al., 1986. III-25

FIGURE III-4 Relationship of Blocrl Lead Level to Height in Children Aged Oto 7

ADJUSTED HEIGHT 1cm) 106.25 • • 108.00 l • • • • • 107.50 •• •• 0 107.00 • • 106.50 • 10600 •

106.50

105.00 4--....._...,_...,__._...... ,, ...... ~.__,_~~_,_...._____._~_.___.___...... ~,_.___.__._...__.__...J.._._-l.__.__.•~ 5 10 15 20 25 30 35 ADJUSTED BLOOD LEAD (..,dll Adjusted height and adjusted blood lead levels for children aged 7 years and younger in Second National Health and Nutrition Examination Survey. Both height and blood lead level have been adjusted by regression for effects of age, race, sex, and all other variables significant at .05 level. Each point is mean height and mean blood lead level of approximately 100 consecutive observations, ordered by blood lead levels. Regression line reflects slope of coefficient obtained from multiple regression analysis of all 2,695 obser­ vations.

$,ource; $chwartz et al., 1986. III-26

At the average age (59 months), the mean blood-lead level of the children appears to be associated with a reduction of about

1.5 percent in the height that would be expected if their blood lead level was zero. The relative impact on weight and chest circumference is of the same magnitude.

III.A.4.c. Summary of Stature Effects The inverse correlation of blood lead and growth in U.S. children is often understood in the context that blood lead is a composite factor for genetic, ethnic, nutritional, environmental, and socio-cultural factors that are insufficiently delineated by age, race, sex and nutrition or by family income, urban residence, and all other available n~tritional indices. An environment that favors a higher blood lead* may supercede all of the established predictors such as socioeconomic status and other demographic characteristics. Growth is a complicated phenomenon, accompanied by an orderly sequence of maturational changes. There are many mechanisms that may account for lead's effect on physical growth and development. Prenatal exposure has an inverse effect on gestational age, which

* Assessments of the risk of ambient lead exposure recognize the triple jeopardy of the urban poor: 1) the exposure to lead from multiple sources is highest in low income areas: 2) in high lead environments, the amount ingested increases with deficiencies in child care and household cleanliness: and 3) the intestinal absorption of lead increases with nutritional deficits. The interaction of socio-cultural and nutritional deprivation with both environmental exposure and absorption of lead has long confounded the delineation of the threshold for behavioral and cognitive effects of low-level lead exposure. III-27 in turn can adversely affect growth. There are known negative interactions between lead and calcium messengers, heme-dependent enzymes, and neuroendocrine function. While the effect is cle~rly plausible, little research has investigated potential mechanisms directly. At least one very recent article (Huseman et al., 1987) uses a rat pituitary model to support the biological plausibility of a neuroendocrine effect on growth. At 20. ug/dl, vitamin D metabolism is potentially sufficiently disrupted to hamper the uptake and utilization of calcium, and children are one-and-a-half times more likely to exhibit abnormal red blood cell indices than at 10 ug/dl. For this analysis, we have assumed that children with blood lead levels over 20 ug/dl are at risk of suffering from s~aller stature. To assess the benefit of this potential rule, we used the NHANES data on the distribution of blood lead levels in the country to calculate the number of children who would be brought below 20 ug/dl of blood lead at an MCL of 20 ug/1: 82,000.

We have ascribed no monetary value to this health effect because it is difficult to put a monetary value on gestational age, fetal development, and children's growth and stature. The co.rrelation with blood lead level is independent of the significant effects on growth of sex, race, and nutritional status, as well as all identifiable measures of socioeconomic status. As yet, it has exhibited no threshold. Common sense, however, suggests the value would be high~

The method used to calculate the number of fetuses at risk of exposure to potentially dangerous levels of lead is described in Section C, below. III-28

III.B. Neurotoxic Effects of Lead Exposure

Lead has been known to be a neurotoxicant since the early

1800s, and neurotoxicity is among the more severe consequences of lead exposure. At very high blood-lead levels, encephalopathy and severe neurotoxic effects are well documented; the neurotoxic effects at lower blood-lead levels, however, are less clearly defined. Recent research has investigated the occurrence of overt signs and symptoms of neurotoxicity and the manifestation of more subtle indications of altered neurological functions in individuals who do not show obvious signs of lead poisoning.

This section presents new data on cognitive effects at low levels of lead exposure. These studies were not discussed in

The Costs and Benefits of Reducing Lead in Gasoline.

III.B.l. Neurotoxicity at Elevated Blood-Lead Levels

Very high blood-lead levels (i.e., above 80 ug/dl in children) are associated with massive neurotoxic effects that can include severe, irreversible brain damage; ataxia (i.e., the inability to coordinate voluntary muscular movements); persistent vomiting; lethargy; stupor; convulsions; coma; and sometimes death. Once encephalopathy occurs, the risk of death for children is significant (Ennis and Harrison, 1950; Agerty, 1952; Lewis et al., 1955), regardless of the quality of the medical treatment they receive. III-29

In cases of severe or prolonged nonfatal episodes of lead encephalopathy, the neurological damage is qualitatively similar to that often seen following traumatic or infectious cerebral injury, with permanent and irreversible damage being more common in children than adults (Mellins and Jenkins, 1955;

Chisolm, 1956, 1968). 'I'he most severe effects are cortical atrophy, hydrocephalus (an abnormal increase in cranial fluid), convulsive seizures, and severe mental retardation. Permanent central nervous system damage almost always occurs in children who survive acute lead encephalopathy and are re-exposed to lead (Chisolm and Harrison, 1956). Even if their blood lead levels are kept fairly low, 25-50 percent show severe permanent sequelae including seizures, nervous disorders, blindness, and hemiparesis (paralysis of half of the body) (Chisolm and Barltrop,

1979).

Even children without obvious signs of acute lead encephalopathy have exhibited persisting neurological damage. As early as 1943, Byers and Lord's study of 20 previously ·1ead­ poisoned children indicated that 19 later performed unsatis­ factorily in school, "presumably due to sensorimotor deficits, short attention span, and behavioral disorders". Effects such as mental retardation, seizures, cerebral palsy, optic atrophy, sensorimotor deficits, visual-perceptual problems, and behavior disorders have been documented extensively in children following overt lead intoxication or even just known high exposures to lead (e.g., Chisolm and Harrison, 1956; Cohen and Ahrens, 1959;

Perlstein and Attala, 1966). III-30

The extent of the later manifestations seems to relate to the severity of the earlier observed symptoms. In Perlstein and Attala, 9 percent of the children studied, none of whom appeared to have severe symptoms when diagnosed for overt lead poisoning, were later observed to be minimally mentally retarded and 37 percent showed some lasting neurological sequelae.

At somewhat lower blood-lead levels (i.e., 30-70 ug/dl), substantial data confirm that a variety of neural dysfunctions occur in apparently asymptomatic children. Several studies indicate that blood lead levels of 50-70 ug/dl are associated with IQ decrements of 5 points. Adverse electrophysiological effects, including markedly abnormal EEG patterns, slow-wave voltages, etc., are also well documented at levels of 30-70 ug/dl and even beiow.

De la Burde and Choate (1972, 1975) showed persisting neuro­ behavioral deficits in children exposed to moderate-to-high levels of lead~ most of the children appear to have had blood lead levels > 50 ug/dl. Compared to low-lead control children -- matched for age, sex, race, parents' socioeconomic status, housing density, mother's IQ, number of children in the family below age 6, presence of father in the home, and mother working the higher lead children averaged about five points lower in IQ and were seven times more likely to have repeated grades in school or to have been referred to school psychologists. Moreover, follow-up studies showed that these effects persisted for at least three years.

The 5-point IQ decrement found in asymptomatic children with blood lead levels> 50 ug/dl is consistent with other studies. III-31

These include Rurnmo {1974) and Rurnmo et al. {1979), which found a

16-point decrement in children> 80 ug/dl and a 5-point loss in

asymptomatic children averaging about 68 ug/dl, and reanalysis

by Ernhart of the data in Perino and Ernhart (1974) and Ernhart et

al. (1981), as described in the Criteria Document (US-EPA, 1986;

p. 12-81 to 12-85).*

While chelation therapy may mitigate some of these persisting

effects, significant permanent neurological and cognitive damage

results from very high lead levels, with or without encephalopathy.

In addition, these children also appear more likely to experience

neurological and behavioral.impairments later in childhood.

III.B.2. Neurotoxicity at Lower Blood-Lead Levels

The adverse effects of lead on neurological functioning, both

on the microscopic {e.g., cellular and enzymatic) level and the

macroscopic (e.g., learning behavior) level, are well ·documented.

On the micro-level, data. from experimental animal studies suggest

several possible mechanisms for the induction of neural effects,

including: (1) increased accumulation of ALA in the brain as a

consequence of lead-induced impaired heme synthesis, (2) altered

ionic balances and movement of ions across axonal membranes and

at nerve terminals during the initiation or conduction of nerve

impulses due to lead-induced effects on the metabolism or synaptic

utilization of calcium, and (3) lead-induced effects on the

metabolism or synaptic utilization of various neurotransmitters.

·* Ernhart submitted the re-analysis, with better control for confounding variables and with errors corrected, to EPA's Expert Committee on Pediatric Neurobehavioral Evaluations {1983). III-32

In addition, lead-induced heme synthesis impairment, resulting in reduced cytochrome C levels in brain cells during crucial develop­ mental periods, has been clearly associated with the delayed develop­ ment of certain neuronal components and systems in the brains of experimental animals (Holtzman and Shen Hsu, 1976). (Cytochrome C is a link in the mitochondrial electron transport chain that pro­ duces energy, in the form of adenosine triphosphate (ATP), for the entire cell.) Given the high energy demands of neurons, selective damage to the nervous system seems plausible.

In addition to the effects of lead on the brain and central nervous system, there is evidence that peripheral nerves are affected as well. Silbergeld and Adler (1978) have noted lead­ induced blockage of neurotransmitter (acetylcholine) release in peripheral nerves in rats, a possible result of lead's disruption of the transport of .calcium across cellular membranes. This disruption of cellular calcium transport may also contribute to the effects of lead on peripheral nerve conduction velocity.

Landrigan et al. (1976) have noted a significant correlation between blood lead and decreasing nerve conduction velocity in children in a smelter community. This effect may indicate advan­ cing peripheral neuropathy.

Paralleling these cellular or biochemical effects are electrophysiological changes indicating the perturbation of peri­ pheral and central nervous. ~ystem fu.nctioning observed in children with blood lead levels of 15 ug/dl and even below. These include slowed nerve conduction velocities (Landrigan et al., 1976), reaction-time and reaction-behavior deficits (Winneke et al., 1984; III-33

Yule, 1984), as well as persistent abnormal EEG patterns including altered brain stem and auditory evoked potentials down to 15 ug/dl

(Benignus et al., 1981; Otto et al., 1981, 1982, 1984). Neuro­ logical effects of lead at such low levels are particularly important because two- and five-year follow-up studies (Otto et al., 1982, 1984) indicated some persistent effects.

A recent article (Schwartz and Otto, 1987) has confirmed earlier findings of hearing effects related to low and moderate lead exposure. This study also found lead levels significantly related to major neurological milestones in early childhood development, including the age at which a child first sat up, walked or spoke. No threshold was evident for either the hearing or developmental effects.

Animal studies have also noted aberrant learning behavior at lower blood-lead levels. Crofton et al. (1980) found that the development of exploratory behavior by rat pups exposed to lead i~ utero lagged behind that of control rats. Average blood-lead levels on the 21st postnatal day were 14.5 ug/dl* for the exposed pups and 4.8 ug/dl* for the controls.

Gross-Selbeck and Gross-Selbeck (1981) found alterations in the operant behavior of adult rats after prenatal exposure to lead via mothers whose blood lead levels averaged 20.5 ug/di.* At the time of testing (3 to 4 months, postnatal), the lead-exposed animals' blood-lead levels averaged 4.55 ug/dl* compared to 3.68 ug/dl* in the controls. This suggests that changes in central nervous system function may persist for months after the cessation

* Blood lead levels in animals are not comparable to human blood­ lead concentrations. III-34 of exposure to relatively low blood-lead levels. In addition, animal studies show that behavior effects include both reduced performance on complex learning problems and signs of hyperactivity and excessive response to negative reinforcement (Winneke, 1977 and 1982a). Finally, these effects show signs of a dose-response relation­ ship. In children with high level lead poisoning, neurological damage is indisputable and mental retardation is a common outcome. For children with somewhat lower blood-lead levels, de la Burde and Choate (1972, 1975) found lesser but still significant cognitive effects, including lower mean IQs and reduced attention spans.

Several studies have found smaller effects at lower blood-lead

levels. Some very recent studies have also shown previously­ undetected, significant cognitive effects in poor black children in the normal range of blood lead levels (from 6 ug/dl) without exhibiting an evident threshold.

While some of these effects have only been observed at higher blood-lead levels, in animals, or in vitro, they show a consistent dose-dependent interference with normal neurological functioning. Furthermore, several of these effects have been documented to occur at very low blood-lead levels (< 10 ug/dl) in children, with no clear threshold yet evident.

III.B.2.a. Cognitive Effects of Lower Blood-Lead Levels

The earliest study of cognitive effects from relatively low levels of lead exposure was done by Needleman et al. (1979), using shed deciduous teeth from over 2,000 children to index lead exposure. Among other findings, this study showed evidence of a I_II-35 continuous dose-response function relating lead levels to behavior as rated by the teacher in terms of attention disorders. The authors also divided the subjects into a high- and low-lead group; significant effects (p < 0.05) were reported for various IQ indices, for classroom behavior, and for several experimental measures of perceptual-motor ability. Numerous papers by Needleman and his co-workers have provided additional analyses and follow-up studies related to the original data. These are listed in the bibliography of this document and are summarized in the Criteria

Document (1986; p. 12-85 ff).

There were many questions relating to the Needleman analysis.

An Expert Committee on Pediatric Neurobehavioral Evaluations, convened by EPA, noted some methodological problems and asked for a reanalysis and some additional analyses (Expert Committee, 1983). Reanalyses were conducted by Needleman (1984), Needleman et al. (1985) and EPA. All of the reanalyses confirmed the published findings of significant associations between lead levels and IQ decrements. After controlling for confounding variables, the Needleman data show evidence of a 4 IQ-point decrement associated with blood levels of 30-50 ug/dl. This finding is consistent with earlier studies showing IQ decrements of 5 points or higher in children with blood lead levels> 50 ug/dl discussed previously (de la Burde and Choate, 1972 and

1975; Rummo, 1974; Rummo et al., 1979; etc.).

Since Needleman's original study in 1979, many other analyses of cognitive effects related to low and moderate lead exposure have been published. While some of these studies, like the earlier III-36 ones, suffer from some methodological flaws and the relatively small sample sizes in many of them made it difficult to prove a statistically significant effect,* the combined weight of the studies point to "various types of neural dysfunction in apparently asymptomatic children across a broad range of blood lead levels"

(Criteria Document, 1986; p. 12-156), including small IQ decre­ ments in children with blood lead levels of 15-30 ug/dl. In addition, new studies have examined neurotoxic effects in younger children and infants. Section A (above) discussed results from studies that found an inverse association between blood lead levels and gestational age at blood lead levels found commonly in the general population. Because gestational age can affect mental development in infants, whatever mechanism lies behind that effect must be factored intG the discussion of lead's neuro­ logical effects.

* A statistical method for combining comparable studies to overcome the problem of small sample size is described in The Costs and Benefits of Reducing Lead in Gasoline (EPA, 1985-,-,­ P• IV-33 ff. For use in public policy making, rejecting the results of studies simply because they fail to attain significance at the 5 percent level may be inappropriate for two reasons. First, policy makers need to be concerned about both type I and type II errors. Significance tests guard only.against the first type (falsely rejecting the null hypothesis of no effect); they help ensure that a regulation is not imposed when there is no adverse effect. Type II errors (failing to reject the null hypothesis when it is false) also can be costly, however, because they can result in the underregulation of a real hazard. With small sample sizes and subtle effects, the probability of a type II error can be large.

The second reason for caution in rejecting the results of these studies, particularly in this case, is that while several fail to attain statistical significance individually, they do show a consistent pattern: the children in the higher lead groups generally have lower mean IQs. III-37

Several studies assessed the relationship between blood lead levels at birth and the subsequent mental development of the infants. Bellinger et al. (1984), observing 216 middle- and upper-middle class children aged 6 months to 2 years in a pros­ pective longitudinal study of early developmental effects of lead, found scores on the Bayley Mental Development Index (MDI) inversely related to umbilical cord blood-lead levels. The subjects were divided into three groups with mean blood-lead levels of 1.8 ug/dl (the low group), 6.5 ug/dl (midgroup), and

14.6 ug/dl (high). Gestational age and some other variables were identified as confounders of the association between cord blood lead and the MDI;* these confounding (positive) associations reduce the degree of association between cord blood lead and the

MDI. Adjusting for confounding, and controlling for all known relevant factors, the difference in scores between the high and low blood-lead level groups was about 6 points on the MDI.

Follow-up studies (Bellinger et al., 1985; 1986a; 1986b) indicated that the association between cord blood-lead level and MDI score continued for at least two years; no association was found with post-natal blood-lead level.

Vimpani et al. (1985), in a longitudinal study of almost 600 children at age 24 months, also found a statistically significant relationship between blood lead levels in infants and their per­ formance on the MDI. Ernhart et al. (1985 and 1986) investigated prenatal lead exposure and post-natal neurobehavioral function,

* That is because gestational age is also related to cord blood lead levels. III-38 as well. Of the 17 neurobehavioral measures examined in the

Ernhart studies, three showed significant relationships to blood lead levels.

Interim results of a longitudinal study presented by Dietrich et al. (1986), observing inner-city children in Cincinnati, showed evidence of an inverse relation between blood lead levels at three months with performance on three major mental development indexes, including the MDI. But these interim findings showed an asspciation only for white infants. This and other analyses have also shown indirect effects on mental development and performance through lead's effect on gestational age and/or birth weight (cf.

Addendum to the Criteria Document, 1986; p. A-35f).

In addition, several studies show an association between blood lead levels and other neurobehavioral patterns. Ernhart

(1985 and 1986) showed that prenatal lead exposure correlated with certain neonatal behavior such as jitteriness and hyper­ sensitivity, as measured on the Neurological Soft Signs scale. A follow-up study (Wolf et al., 1985) showed evidence that lowered

Bayley MDI scores for one year olds was a sequela of the cord blood-lead relation shown on the Neurological Soft Signs scale after birth. And Winneke et al. (1985a) showed a significant relationship between perinatal blood-lead levels and one measure of psycho-motor ability at ages 6-7.

Finally, two new general population studies (Schroeder et al.,

1985; and Schroeder and Hawk, 1986) investigated low socio-economic status children with blood lead levels in line with (or just slightly higher than) levels in the general population, controlling III-39

for socio-economic factors, age, race, etc. The first study examined 104 lower SES children with blood lead levels ranging from 6-59 ug/dl (mean about 30 ug/dl). This study found a signi­ ficant effect (p < 0.01) of lead upon IQ, which was sufficient to disrupt the normal mother-child IQ correlation. The second study, replicating the previous study with 75 low SES black children showed a highly significant relationship (p < 0.0008) between IQ and blood lead levels over the low to moderate range of 6-47 ug/dl. These studies suggest that lower socio-economic status

places children at greater risk of the deleterious effects of

low-level lead exposure on cognitive ability, while confirming

that other factors (maternal IQ, home environment, etc.) are also closely related to IQ.

Winneke et al. (1985a and 1985b) also examined the predictive value of different markers of lead exposure for subsequent neuro­

behavioral development. Of an original study of 383 children at

birth, 114 subjects were followed-up at ages 6-7. Mean blood-lead

levels (maternal and infant) had been 8-9 ug/dl (range: 4-31 ug/dl). Regression analyses showed that maternal blood-lead

levels (related closely to umbilical cord levels) accounted for

nearly as much of the variance in neurobehavioral test scores at age 6-7 as did contemporary blood-lead levels.

The combined results of available studies of cognitive effects

at low and moderate lead levels present evidence of potential IQ

decrements and other cognitive impacts due to lead exposure at blood lead levels found commonly in the U.S. population; i.e., at

15 ug/dl and below, and possibly as low as 6 ug/dl for some groups of children. III-40

III.B.3. The Magnitude of Lead's Impact on IQ

The studies summarized above indicate that among the cognitive effects resulting from exposure to lead is a potential lowering of children's IQs and a reduction in their ability to perform well in school. The latest draft of the Criteria Document {1986) characterizes the evidenc~ as suggesting that, on average, blood lead levels of 50 to 70 ug/dl could correlate with average IQ decrements of five points, blood lead levels of 30 to 50 ug/dl could be associated with a four-point decrement in IQ, and that lead levels of 15 to 30 ug/dl could be related to IQ reductions of one-two points {p. 12-156, 12-282, and elsewhere). In Section

III.D, we monetize the benefits of reducing these effects using the costs of compensatory education and potential decreases in future earnings resulting from decreased IQ.

These levels of effects may be associated with relatively consistent and/or relatively long exposure periods, possibly even several years. Permanent IQ effects may result only from fairly long periods of exposure, and a child who has a certain blood lead level for a relatively short amount of time {perhaps, a few months) may not suffer the full effect. Because of this uncertainty, for the effect upon decreased future earnings, we assumed conservatively that a child must be at a certain lead level for 3-4 years before permanent and irreversible IQ loss results.

This is an extremely conservative assumption. The data on prenatal exposure {obviously limited to at most nine months) show significant effects {both physical and neurological) persisting for two years or more. Fetal development is arguably particularly III-41 vulnerable to disruption~ Nonetheless, much data suggest that post-natal exposure periods of a year or so produce detectable effects. In addition, even if a lead-induced lag in cognitive or physical development were no longer detectable at a later age, this does not necessarily mean that the earlier impairment was without consequence. Given the complex interactions that contri­ bute to the cognitive, emotional, and social development of children, compensations in one area of a child's development may exact a cost in another area. Unfortunately, very little is known about how to accurately measure these interdependencies.

We have chosen the conservative estimate of 3-4 years of exposure because of unanswered questions of reversibility and permanence.

The approach of ascribing benefits only to those children who are brought below a critical threshold (15 ug/dl, 30 ug/dl or

50 ug/dl) by this proposed rule suffers from several faults, which cut in opposite directions and may offset each other.

Categorization does not account for the fact that some children who are prevented by the regulation from going over a given

threshold will do so by a narrow margin {e.g., their blood lead

level will be 29 ug/dl when it would have been 31 ug/dl in the

absence of the rule); such children are unlikely to receive the full four-point gain in IQ but they will receive more than one point. This means that the benefits of this potential rule may

be overestimated. On the other hand, categorization attributes no benefit to

children whose blood lead levels are reduced from very high levels, III-42 but not brought below a given threshold or to those whose levels would have been between two thresholds without the rule, but whose levels decrease further by the reduction in lead in drinking water. Also, it is quite possible that children suffer long­ lasting, even permanent, effects with shorter exposure periods than the 3-4 years we assumed. These factors indicate that our benefit estimate may be too low.

III.C. Fetal Effects

Lead's adverse effects upon human reproductive functions have been known for over 100 years.* In 1860, for instance, Paul published findings (cited in the Criteria Document, p. 12-192) that lead-poisoned women were likely to abort or deliver stillborn

infants. Because lead passes the placental barrier and fetal lead uptake continues throughout development, a growing concern in the public health community is that the most sensitive popula­ tion for lead exposure is fetuses and newborn infants. This concern is supported by both animal and human studies. Several categories of fetal effects were discussed previously.

Within Section A, above, in the discussion of lead's adverse effect upon children's physical growth and development, we presented data on the inverse relationships between blood lead

levels and gestational age, birth weight and birth height. Section B, also above, describes the neurotoxic effects of lead,

including the inverse relationship between blood lead level and

* Indeed, 'lead plasters' were used as abortifacients at the turn of the century. III-43 infant mental development as measured.by several different neurological indices.

In addition, several studies have implicated lead in compli­ cations of pregnancy, including early and still b~rths, and, possibly, low-level congenital anomalies. (Lead's adverse effects upon reproductive function are discussed in Section IV.B. of the next chapter.) As discussed previously, lead has a negative effect upon gestational age. As early examples of these findings,

Fahim et al. (1976) found that women who had normal full-term pregnancies had average blood-lead levels of 14.3 ug/dl, whereas women with early membrane rupture had average blood-lead levels of 25.6 ug/dl, and women with premature delivery had average blood-lead levels of 29.1 ug/dl. Wibberly et al. (1977) found that higher lead levels in placental tissues were associated with various negative pregnancy outcomes, including prematurity, birth malformation, and neonatal death. Bryce-Smith et al. (1977) found bone lead concentrations in still births of 0.4-24.2 parts per million (ppm) in the rib (average: 5.7) versus typical infant bone lead levels of 0.2-0.6 ppm. Needleman et al. (1984) analyzed data from over 4,000 live births at Boston Women's Hospital and reported an association between minor congenital anomalies and umbilical-cord blood-lead levels. There was no association between any particular malfor­ mation and lead, but only between all minor malformations and lead. There also were no significant associations between lead and any major malformations, although given the rate of such malformations in the general population, a sample this size has little power to detect such an effect. Holding other covariates III-44 constant, the relative risk of a child demonstrating a minor malformation at birth increased by 50 percent as lead levels increased from 0.7 ug/dl to 6.3 ug/dl (the mean cord-lead level).

This risk increased an additional 50 percent at 24 ug/dl.

(Umbilical-cord blood-lead levels are generally somewhat lower than, but correspond to, maternal blood-lead levels; e.g., Lauwerys et al., 1978.)

Two other studies (McMichael et al., 1986; Ernhart et al.,

1985 and 1986) investigated the association between pre-natal lead exposure and congenital morphological anomalies. They did not find a similar occurrence of congenital anomalies. On the other hand, the Needleman analysis relies upon a much-larger data base than either the Ernhart or McMichael studies. Nonetheless, the available evidence on lead's effect on congenital anomalies allows no definitive conclusion about low-level lead exposures and the occurrence of congenital anomalies.

Finally, Erickson et al. (1983) found lung- and bone-lead levels in children who died from Sudden Infant Death Syndrome were significantly higher (p < 0.05) than in children who died of other causes, after controlling for age. While this study suggests a potential relationship between lead exposure and

Sudden Infant Death Syndrome, this issue also remains to be more fully evaluated.

III.C.1. Assessing the Benefits of Reduced Fetal Exposure to Lead

Lead crosses the placental barrier and fetal uptake of lead continues throughout development. Fetal and new-born blood-lead III-45 levels are closely related tou though generally slightly lower than, maternal levels.* In addition, during physiological condi­ tions of bone dimineralization, which are known to occur during pregnancy and lactation, lead as well as calcium may be released from its storage in bone. A readily mobile compartment of skeletal lead has been demonstrated in humans {Rabinowitz et al., 1977) and in experimental studies both in vivo (Keller and Doherty, 1980a and 1980b) and in vitro (Rosen, 1983; Pounds and Rosen, 1986).

For these reasons, to assess the benefits of EPA's proposed reduc­ tion in the MCL for lead, we are concerned with two populations: pregnant women currently at risk of receiving high levels of lead in their drinking water and especially those women {aged 15-44) who are likely to have blood lead levels that could present a risk to the unborn child.

To determine what blood lead level should be used as a cut­ off for estimating risk to the fetus from lead exposure, two recent policy actions related to EPA rules were considered. In the Spring of 1985, EPA's Clean Air Science Advisory Commlttee recommended a goal of preventing children's blood-lead levels

* "Exposure levels during the course of pregnancy may not be accurately indexed by blood lead levels at parturition. Various studies indicate that average maternal blood-lead levels during pregnancy may tend to decline, increase, or show no consistent trend. These divergent results may simply reflect the likelihood that the maternal blood lead pool is subject to increase as bone stores of lead are mobilized during pregnancy and to decrease as lead is transferred to the placenta and fetus. Apparently, then, under some condi­ tions the fetus may be exposed to higher levels of lead than indicated by the mother's blood lead concentration." {Text quoted from The Addendum to the Criteria Document, 1986; p. A-45. See references cited there.) III-46 from exceeding 15-20 ug/dl. This goal was supported by the epidemiological and toxicological data at the time that showed significant lead-induced health effects beginning to be detec­ table in that range. More recently, the Addendum to the Criteria Document (US-EPA, 1986; p. A-48) stated, "At present, • . . ' perinatal blood lead levels at least as low as 10 to 15 ug/dl clearly warrant concern for deleterious effects on early post­ natal as well as prenatal development."

To be conservative, we used the intersection of these two ranges -- 15 ug/dl -- as the blood lead level of concern for fetal effects. As the simplest relationship between maternal and fetal blood-lead levels, we have assumed equivalence. This ignores some findings that the rate of fetal absorption of lead may increase throughout development (e.g., Barltrop, 1969;

Rabinowitz and Needleman, 1982; Donald et al., 1986). (See the

Criteria Document, Sections 10.2.4 and 12.6, for a fuller dis­ cussion of this issue.) The Air Quality Criteria Document estimates that there are approximately 54 million women of childbearing age (i.e., between 15 and 44 years old), of whom about 7 percent are likely to be pregnant at any given time (1986; p. 13-47); this is about the same percentage as the annual birth rate given by the Census

Bureau: 67.4 per 1,000 women aged 15-44.* Census Bureau data

(presented in Current Population Reports, Series P-25, or summarized in Table 27 in Statistical Abstracts, 1986) show that 24 percent

* 1985 Statistical Abstracts of the United States (1986), Table 82. III-47 of the total population is women aged 15-44. Assuming that women of childbearing age and that pregnant women are distributed proportionately among those served by community water systems. and by private or non-community water supplies, and assuming that these women are distributed proportionately between community water systems with high and low lead levels, these figures yield the following estimates of fetuses at risk of exposure to lead levels exceeding 20 ug/1 in drinking water.

24% x 42 million* x 67.4 per thousand= 680,000 fetuses at risk

Calculating the number of women at high blood-lead levels is more difficult. Our population models do not yet include suffi­ cient data on all adult women in the United States because our analyses so far have focused, separately, on children and adult men. However, some data are available that enable us to make very crude exposure estimates for this pop~lation. The Hispanic Health and Nutrition Examination Survey (called either the Hispanic HANES or HHANES), conducted by the National

Center for Health Statistics (NCHS) between 1982 and 1984 contains relatively recent data on blood lead levels in the U.S., including adults. These data, published and available from NCHS, indicate

.that in 1988 Mexican American women aged 15-44 will have an estimated mean blood-lead level of 7.1 ug/dl, with a geometric standard deviation of 1.5; they also show that an estimated 0.36

* Estimate of people in the United States served by community water supplies who currently receive water that exceeds 20 ug/1. Methodology presented in Chapter II. III-48 percent of those women will have blood lead levels over 15 ug/dl in 1988.* While we do not know how data on Mexican Americans compare to data on white Americans, these estimates certainly significantly underestimate the lead levels of black Americans, which are generally much higher than whites.** Assuming that all women in the u.s. have blood lead levels and distributions comparable to those in the Hispanic HANES: 219 54 million x 0.36% x 240 million= 177,000 women in 1988 served by community water supplies are likely to have blood lead levels over 15 ug/dl, of whom

177,000 X 7% = 12,400 are likely to be pregnant in any given year. For these women, any contribution of lead from drinking water is a potential health risk to the fetus because they are already at the cut-off point recommended by the Clean Air Science Advisory Committee, and the fetus may take the bulk of the lead absorbed by the mother. Determining what fraction, if any, of these most-at-risk fetuses {i.e., of mothers with blood lead levels> 15 ug/dl and receiving water> 20 ug/1) is included in the 680,000 at-risk

* These estimates rely upon the coefficients on the distribution of blood lead levels in the country from the NHANES II {dis­ cussed previously) and Census Bureau data on the demographics of the population. These extrapolations assume reduced exposure to lead from gasoline.

**Foran indication of the racial differences in the distribution of blood lead levels, look at the data presented on Table III-1, on p. III-8 above. III-49 fetuses (i.e., of mothers receiving water> 20 ug/1) estimated above is difficult. To avoid double-counting we have used the most conservative and least controversial assumption: that all of the most-at-risk fetuses are included in the 680,000 fetuses exposed in utero to drinking water exceeding the proposed MCL.

Therefore, the proposed MCL would protect 680,000 fetuses in 1988.

III.D. Monetized Estimates of Children's Health Benefits

The health benefits of reducing children's exposure to lead are diverse and difficult to estimate quantitatively or to value in monetary terms. The monetized benefits include only two admittedly incomplete measures: savings in expenditures for medical testing and treatment, and savings associated with decreased cognitive ability. These measures of benefit exclude many important factors, such as the reproductive and stature effects discussed above. These and other limitations are discussed in Section III.E, below.

In fact, many children with elevated blood-lead levels are neither detected nor treated. However, this estimation procedure assumes that children who go undetected and untreated bear a burden at least as great as the cost of testing and providing the treatment (medical or educational) of those who are detected.

So, all children with high blood-lead l.evels are assumed to incur

"costs"-, whether medical expenditure costs or personal costs in the form of poor health, inadequate learning, decreased future vocational options, etc. III-50

III.D.l. Reduced Medical Costs

To estimate the benefits of reduced medical care expenses, we used the optimal diagnosis, treatment, and follow-up protocols recommended by Piomelli, Rosen, Chisolm, and Graef in the Journal of Pediatrics (1984). Piomelli et al. estimated the percentages of children at different blood-lead levels who would require various types of treatment. Figure III-4 summarizes the treatment options that we used, based on the recommendations of Piomelli et al.

An evaluation of typical medical services suggests that administrative expenditures and follow-up tests would cost $110 (1985 dollars) for eaeh child found to be over 25 ug/dl at screen­

ing. Of those ehildren over 25 ug/dl blood lead, based on Piomelli et al. (1982) and Mahaffey et al. (1982), we estimated that 70 percent would be over 35 ug/dl EP. Piomelli et al. (1984) recommend provocative ethylenediamine-tetraaeetic aeid (EDTA) testing for such children. EDTA testing typically requires a day in the hospital and a physician's visit; based upon hospital cost data, we assessed a cost of $540 per test (1985 dollars). We also assumed that all ehildren receiving EDTA testing would receive a series of follow-up tests and physicians' visits, costing an estimated $330 (1985 dollars).

The purpose of EDTA testing is to see if children have a dangerously high body-lead burden (a lead excretion ratio over 0.60, per Piomelli et al.). Table III-2 presents Piomelli et al.'s estimates of the percentages of children at various blood lead levels who will require .ehelation therapy; it ranges from a III-51

FIGURE III-5. Flc,.., Diagram of Medical Protocols for Children with Blood Lead Levels above 25 ug/dl

Blood Lead Levels >25 ug/dl I elevated FEP?

no yes .,

high body sinple lead b.lrden?* follc,..,-up no yes I I chelation long therapy foll0,,1-up I repeat chelation? no yes I long chelation follow-up therapy

I repeat NarES: chelation?

*Provocative IDI'A or no yes other test I tChelation therapy, because long chelation of its severe side-effects follc,..,-up therapy and inherent dangers, cannot be rEpeated again after this I point long follc,..,-upt III-52

TABLE III-2. Percent of Children Requiring Chelation Therapy

Blood Lead Levels Percent

25-30 ug/dl 0* 30-39 ug/dl age three and over 9.6* age under three 11.5*

40-49 ug/dl age three and over 26.0 age under three 37.9

50-59 ug/dl age three and over 36.0 age under three 49.0 above 59 ug/dl 100.0

Source: Piomelli et al., 1984

* At blood lead concentrations of 25 to 35 ug/dl, 6-7 percent of children require chelation therapy (Piomelli et al., 1984). The presentation above, while consistent with the data and other presentations, can easily result in an underestimation of risks.

In addition, given the degree of non-linearity between blood lead levels and ehelatable lead, it is possible that some children between 15 and 24 ug/dl (a concentration below CDC's definition of 'elevated' blood lead) may have positive EDTA test results. Thus, the percentages above may represent conservative estimates. (Based upon data and discussions with Dr. John Rosen, Department of Pediatrics, Montefiore Medical Center, Albert Einstein College of Medicine; one of the authors of the Piomelli article.) III-53

low of zero for those under 30 ug/dl to a high of 100 percent for those over 59 ug/dl.

Based on the data from the NHANES II, we estimated that, of those children over 25 ug/dl blood lead, about 20 percent are between 30 and 40 ug/dl and 10 percent are over 40 ug/dl. Using those estimates and the percentages in Table III-2, 5 percent of

the children above 25 ug/dl would require chelation therapy.

In addition, we estimated that half of those children chelated would require a second chelation due to a rebound in their blood

lead level, and that half of those children would require a

third chelation treatment. Thus, a total of 0.0875 chelations would be required for every child over 25 ug/dl blood lead at

screening. To calculate the total eost per chelation, we estimated

that it would require five days in the hospital, several physicians'

visits, laboratory work, and a neuropsychologieal evaluation, for a total eost of about $2,700 per chelation.

Multiplying each of these costs by its associated probability and then summing them yields the estimated cost per child over

25 ug/dl:

1.0($1~0) + 0.7($540) + 0.7($330) + 0.0875($2,700)= $955.25, which we round to $950 in 1985 dollars.

Because we have not included welfare losses (such as work

time lost by parents), the adverse health effects of chelation therapy itself (such as the removal of necessary minerals and

potential severe kidney damage), or such non-quantifiables as the III-54 pain from the treatment, this estimate of the benefits is conser­ vative. As mentioned previously, these medical costs are a measure of avoidable damage for all the incremental cases of lead toxicity, whether detected or not.

III.D.2. Costs Associated with Cognitive Damage

The studies on the neurotoxicity of lead show a continuum of effects, in a dose-response relationship, from low to high levels of exposure. Manifestations of this neurotoxicity are varied and include IQ deficits and other cognitive effects, hearing decre­ ments, behavioral problems, learning disorders, and slowed neuro­ logical development. Several of lead's neurotoxic effects can combine, and a few studies (e.g., de la Burde and Choate, 1972,

1975) show poorer performance in school associated with higher blood-lead levels. For instance, these studies showed that children with higher blood-lead levels were seven times more likely than similar children with lower lead levels to repeat a grade, to be referred for psychological counseling, or to show other signs of significant behavioral effect. Supplementary educational programs may compensate for some of these effects, though certainly not all of them.

Because of the difficulties inherent in monetizing neuro­ logical effects, we selected only one sub-category to investi­ gate further -- cognitive damage resulting from exposure to lead.

We developed two methods for calculating the benefits of reduced cognitive damage: compensatory education costs (as a proxy for the damage) and decreased future earnings as a function of IQ points lost. III-55 rrr.D.2.a. Compensatory Education

To estimate roughly the cost of compensatory education for children suffering low-level cognitive damage, we used data

from a study prepared for the Department of Education's Office of

Special Education Programs. Kakalik et al. (1981) estimated that part-time special education for children who remained in regular classrooms cost $3,064 extra per child per year in 1978: adjust­

ing for changes in the GNP price deflater yields an estimate of

$4,640 in 1985 dollars. This figure is quite close to Provenzano's

(1980) estimate of the special education costs for non-retarded, lead exposed children.

In developing ·the algorithm for calculating a unit cost for compensatory education, we made three relatively conservative assumptions. First, we assumed that no children with blood lead levels below 25 ug/dl would require it. This is conservative because many studies show detectable cognitive effects at 15 ug/dl.

Second, we assumed that only 20 percent of the children above 25 ug/dl would be severely enough affected to require and receive

some compensatory education. Third, based upon several follow-up studies that showed cognitive damage persists for three years or more (even after blood lead levels have been lowered), we assumed that each child who needed compensatory education would require it for three years but that the damage would then be compensated

for. This is conservative for two reasons. As a neurotoxin, lead affects many capacities: hearing, motor coordination and other sensory perceptions, as well as cognitive abilities. No part-time in class special education can possibly compensate for III-56 all these effects. In addition, while no studies have yet been published on this, data from several large lead poisoning preven­ tion programs (records on about 1,600 children treated at the

Montefiore Center in New York) show that once lead toxic children require compensatory education, such educational services generally will be required for many more than three years. We have used the three-year cut-off because the only published follow-up studies show cognitive effects ~ersisting for (at least) that period of time. Thus, the estimated average annual cost per child over 25 ug/dl is

(0.20) X (3) X ($4,640) = $2,7~4, which we round to $2,800 (1985 dollars), for compensatory education to address lead's cognitive damage.

III.D.2.b. Effect Upon Future Earnings

Literature concerning the economic returns of schooling has included some investigation of the impact of IQ upon earnings; a survey of this literature was prepared for EPA by ICF (ICF, 1984) and peer reviewed by a panel of distinguished economists. Typi­ cally, estimates of the returns to schooling are based upon an "earnings capacity" that consists of equations for schooling,

"ability" (usually measured by scores on standardized IQ tests), and socioeconomic variables. Both the main subject (economic returns of schooling) and its off-shoot (earnings as a function of IQ) are extremely complicated and controversial.

Despite the wide variety of data sets and methodologies used to examine these issues, the estimates of the direct effect of III-57 ability on earnings appear to be fairly consistent. With one exception, estimates of the direct effects of a one point change in IQ fall between 0.20 and 0.75 percent of future expected earn­ ings. There are fewer estimates of the indirect effects of IQ on future earnings (which include the impact of IQ on the schooling of the child, which in turn affects expected earnings) and they range from 0.18 to 0.56 percent per IQ point. Combined, a one

IQ-point change is associated with a change of 0.65 to 1.15 percent in earnings. We used the arithmetic mean (one IQ point=

0.90 percent of earnings) to calculate the benefit of this rule.

As summarized above (Section III.B.), the literature indi­ cates that children with blood lead levels between 15 and 30 ug/dl could suffer IQ losses of 1-2 points (for which we used the arith­ metic mean -- 1.5 points -- as the point estimate), between 30 and

50 ug/dl children could lose 4 IQ points, and over 50 ug/dl they could lose 5 IQ points. Because permanent IQ damage probably occurs after a year or more of lead exposure, we assumed conserva­ tively that children would suffer these losses after 3-4 years of exposure at these levels.* To calculate the annual benefits of this proposed rule, therefore, the potential effect upon future earnings resulting from these exposure levels was divided by 3.5 years. Multiplied together, reducing a child's blood-lead,level below 15 ug/dl could increase expected future lifetime earnings by 0.4 ~ercent (0.9% x 1.5 IQ points~ 3.5 years= 0.4%),

* Because recent data is showing that much shorter exposure periods produce effects that can last for at least 2-3 years, subsequent analyses will have to re-examine this assumption. III-58 reducing a ehild below 30 ug/dl could increase earnings by 1.0 percent (0.9% x 4 IQ points~ 3.5 years= 1.0%), and reducing a ehild below 50 ug/dl could increase earnings by

1.3 percent (0.9% x 5 IQ points~ 3.5 years= 1.3%).

The present value of expected lifetime earnings is $687,1501* deferred for 20 years at 5 percent real discount rate** reduces it to $258,950 (1985 dollars). The effeet of the cognitive damage would decrease expected future earnings by $1,040 (1985 dollars) for a child brought below 15 ug/dl1 $2,600 (1985 dollars) for a child brought below 30 ug/dl1 and $3,350 (1985 dollars) for a child brought below 50 ug/dl.

To calculate the annual benefits in this category, the number of children who would be brought below each of these points (15, 30 and 50 ug/dl) was multiplied by the ehange in expected future lifetime earnings. Table III-3 presents those

benefit calculations for the proposed MCL reduction to 20 ug/1

for sample year 1988.

* Caleulated from Bureau of the Census data:, Lifetime Earnings Estimates for Men and Women in the United States: 1979 (1983) -- p.3 -- and 1985 Statistical Abstracts of the United States (1986) -- Table 216. Converted to 1985 dollars. ** These costs are deferred because those suffering the effe6ts ar~ children and will not enter the work force for up to 20 years. Obviously, using the largest deferral period (20 years) reduces, the value of the benefit and reduces the benefit es½imate, whereas 8- or 10-year-old children may begin working within 8 years and so would have a much shorter deferral period. This biases the estimates downward slightly. III_.59

TABLE III-3. Estimated Annual Benefits of Reduced IQ Damage by Using Changes in Expected Future Lifetime Earnings, For Sample Year 1988

Blood Lead Level

15 ug/dl 30 ug/dl 50 ug/dl TOTAL

Number of ehildren 230,000 11,000 100 241,.100 IQ points 1-2 4 5 NA potentially lost per child per child per child

Present value of :n ,040 $2,600 $3,350 decreased earnings per ehild per child per child NA · ( 1985 dollars)

TOTAL $239.2 $28.6 $ 0.3 · (1985 $268.1 dollars) million million million million III-60

III.D.3. Summary of Monetized Benefits

Table III-4 summarizes the estimates of the monetized annual children's health benefits of the potential rule for one sample year: 1988. Adding the estimates of compensatory education

($2,800) and medical costs ($950) yields a combined annual benefit estimate of $3,750 per case avoided of a ehild's blood-lead level exceeding 25 ug/dl. This is Method 1 in Table III-4. Method 2 is medical expenses plus decreased future earnings as a consequence of lead's adverse effect upon IQ. The benefits of avoided cogni­ tive damage calculated as a function of IQ's relationship to expected future earnings are not linear, however; they are a step function.* Therefore, average costs per child were not calculated. Instead, we calculated the annualized benefit from avoided IQ losses of reducing 230,000 children below 15 ug/dl; 11,000 children below 30 ug/dl; and 100 below 50 ug/dl, and combined that with the total medical expenses avoided for bringing 29,000 children below 25 ug/dl. Note that the difference between Method 1 and Method 2 is that they include alternative methods for valuing aspects of the cognitive damage resulting from exposure to lead.**

The benefits for Method 1 (medical eosts plus compensatory education) are not absolutely comparable to those from Method 2, for three reasons. First, Method 1 based upon per child estimates -- is strictly a function of the number of children

* Measurements were taken for the step function at 15 ug/dl, 30 ug/dl, and 50 ug/dl. ** This biases the results downward because there is a strong rationale for considering these effects as additive. com­ pensatory education is unlikely to fully compensate for the neurological damage caused by exposure to lead and so, effects upon future ability and performance could still result in decreased earnings. III-61

TABLE III-4. Monetized Annual Benefits of Redueing Children's Exposure to Lead Using Alternative Methods, for Sample Year 1988 (1985 dollars) METHOD 1 METHOD 2 Medical Canpensatory Total Medieal Earnings Total Expenses. Edueation Costs Method 1 Expenses Lost Method 2 Number of children 29,000 29,000 29,000 29,000 241,100 241,100

Costing $950 $2,800 $3,750 $950 $268.1 NA unit per ehild per ehild per ehild per child million total Total $27.6 $81o2 $108.8 $27.6 $268.1 $259.7 benefits million milHon million million million million III-62 passing one critical point: 25 ug/dl. Method 2, on the other hand, depends upon the distribution of children by blood lead levels, and how the distribution changes* as a result of EPA's proposed regulatory action. Second, Method 1 ascribes no benefit to any health effects below 25 ug/dl, while Method 2 includes neurotoxic effects between 15 and 25 ug/dl. Finally, Method 2 measures effects to a child over a working lif~ for each year of exposure, while Method 1 includes only costs incurred for those children who receive compensatory education for the duration of that benefit.

In addition, while they are discussed qualitatively, no monetary value is assigned to the fetal effects, the increased risk of anemia, metabolic changes or the negative impact upon stature. So the monetized benefit estimates omit many important categories, and thus are likely to be significant underestimates of the total benefits of reducing lead in drinking-water. The next section contains a discussion of some of these factors.

III.E. Valuing Health Effects: Caveats and Limitations

To begin valuing the health effects that would be avoided as a result of the new MCL for lead in drinking water, we estimated 1) medical treatment and monitoring costs for those children whose blood lead levels reach or exceed the criteria recommended by the Centers for Disease Control as determining lead toxicity

* That is, how many children pass each of several critical points, depending upon the category: 15 ug/dl, 25 ug/dl, 30 ug/dl or 50 ug/dl. III-63

(25 ug/dl of blood lead when combined with erythrocyte proto­ porphyrin levels of 35 ug/dl), and 2) two alternative ways of valuing the cognitive damage resulting from lead exposure: the cost of part-time in-class compensatory instruction as a proxy for the cognitive damage that lead causes and decreased expected future earnings as a function of IQ loss.

The cost-of-illness estimates themselves are low, primarily

f because, to reduce potential controversy, the calculations rely upon many conservative assumptions. For instance, the monetization of compensatory education costs is based upon likely practice and not preferred treatment. Although children suffering cognitive damage from lead exposure should receive more intensive and extensive educational resources, the Department of Education estimated that they would probably receive only part-time in- class remedial/compensatory help. In addition, the estimate that only 20 percent of children over 25 ug/dl would receive any extra help is conservative. The real (social) cost of the illness does not decrease if not all victims receive the treatment they need; assuming the treatments are efficacious, children who are left with diminished cognitive abilities incur a cost at least equal to the cost of the treatment they should have (but did not) receive. The health benefit estimates, therefore, should be understood as very low lower-bounds for these categories of effects. We have also not conducted 9ost-of-illness calculations for most of the adverse health effects associated with human exposure III-64

to lead. Among the many effects not valued monetarily in the -health benefits analysis are: kidney effects, detectable in children at about 10 ug/dl; hematopoietic damage, detectable in children at below 10 ug/dl; neurological effects in children below the level of lead toxicity, with central nervous system effects detectable at below 10 ug/d~ and no perceived threshold; - metabolic changes, detectable in children at about 12 ug/dl; - enzymatic inhibition, with no threshold indicated in children, even below 10 ug/dl; - all effects on fetuses in vitro, although lead crosses the placental barrier and maternal blood-lead levels correlate with adverse pregnancy outcomes, including decreased gestational age, slowed mental and physical development in neonates, potential low-level congenital anomalies and other adverse outcomes, including fetotoxicity at high levels; - stature effects on children, which are dose-dependent with no threshold evident; and effects upon other organ systems, for instance, immune and gastrointestinal. Finally, three serious phenomena of lead's adverse effect upon human health are not included. First, hematopoietic, meta­ bolic, and enzymatic d~mages have cascading effects throughout the body, which are not adequately addressed. Second, many of the specific effects have long-lasting sequelae that are not included. III-65

And last, there is a significantly greater chance of serious effects later in life, including renal failure and neurological disorders, even in individuals whose highest detected blood-lead level was below that associated with the most severe effects and who did not at the time show evidence of lead toxicity: this risk is not included in the analysis.

In addition to the categories of adverse health effects for which we have not yet been able to quantify benefits at all, the costs of the illnesses that are calculated greatly underesti­ mate the real (social) benefits of preventing those effects, even for the health categories evaluated. The underestimates occur because of the exclusion of some categories of direct costs associated with those effects and the total exclusion of all indirect but related costs (e.g., work time lost by the parents of lead-poisoned children).

In general, ~ociety's willingness-to-pay to avoid a given adverse effect is many times greater than the cost of the illness itself, so cost-of-illness analyses inherently underestimate the benefits of avoiding the adverse effect.* Willingness-to-pay studies indicate that society is usually willing to pay two ·to ten times the cost of medical treatment, and that in specific circumstances society is willing to pay a hundred or a thousand times the cost.of the illness itself in order to prevent its occurrence.

* For instance, in general people would be willing to pay more than the price of two aspirins to avoid having a headache. III-66

More specifically, in the cost-of-illness analyses included in this report, only the expenses that are directly related to an individual's medical treatment for the specific symptom being evaluated at the time the symptom occurs were included. So, for instance, no costs are ascribed for the possibility of adverse effects from the medical treatment or hospitalization itself, or for the possibility that the specific effect of lead may precipitate or aggravate other health effects (e.g., children with anemia are more susceptible to many infections). Related expenses, such as the travel costs to obtain medical services or the costs of making the home environments of children suffering from lead toxicity safe for them (i.e., altering their diet-to compensate for their propensity to anemia, removing all lead-laden dust, etc.) were also excluded. Finally, no value was ascribed to the pain and suffering of those affected; this is an especially significant omission because, as examples, chelation therapy is extremely painful and having a child with lead poisoning or who is hospitalized can totally disrupt family life. We have also omitted all the indirect but related costs of lead's adverse effect upon human health. These include work time lost by friends and relatives of the victims (including the parents of lead-poisoned children); medical research related to the prevention, detection, or treatment of the effects of expo­ sure to lead; the development of new procedures to correct the damage resulting from lead exposure; decreased future earnings for those suffering cognitive damage (other than very young III-67 children) or physical incapacitation (including behavioral dis­ orders) from lead's adverse effe~ts upon virtually every human system; and the like.

III.F. Summary of Annual Monetized and Non-monetized Children's Health Benefits

This chapter presents evidence of a variety of physiological effects associated with exposure to lead, ranging from relatively

subtle biochemical changes to severe damage and even death at very high levels. Of these, only two categories of effects are monetized: costs of medical treatment for children with elevated blood-lead levels and costs associated with the cognitive damage resulting from lead's neurotoxicity. For the latter category,

two alternative monetization techniques were presented: compensa­

tory education as a proxy measure and decreased future earnings as as function of IQ points lost. In addition, the numbers of children at risk of several other pathophysiological and neurotoxie damage eaeh year, in0l.uding those at risk of stature decrements, at increased risk of anemia, total number of children at risk of IQ-point-loss, and fetuses exposed to potentially dangerous lead

levels, were estimated. Table III-5 summarizes both the monetized and non-monetized benefits of reducing exposure to lead in drinking water for one sample year, 1988. III-68

TABLE III-5. Surnnary of Annual Monetized and Non-monetized Children's Health Benefits of Reducing Lead in DrinkinJ Water for Sample Year 1988

Annual Monetized Benefits (1985 dollars)

0 Medical costs $27. 6 million ~Cognitive damage costs: canpensatory education (Method 1) $81. 2 million decreased future earnings (Method 2) $268.1 million

'IOTAL Method 1 $108.8 million Method 2 $295. 7 million

Annual Non-monetized Benefits (children at risk of:)

0 Requiring medical treatment 29,000

0 Loss of 1-2 IQ points 230,000 4 IQ points 11,000 -? IQ points 100

0 Requiring compensatory education 29,000

0 Stature decrement 82,000 ~Fetuses at risk 680,000

0 Increased risk of hematological effects 82,000 CHAPTER IV HEALTH BENEFITS OF REDUCING LEAD: ADULT ILLNESSES

Concerns about the health effects of ambient exposure to lead traditionally have focused on children. Although lead has a variety of adverse effects on the health of adults, most of these effects were believed to be a risk only at high blood-lead levels.

Recently, many analyses -- still the subject of some controversy

-- have shown a robust, continuous relationship between blood lead levels and blood pressure in men, confirming a relationship between lead exposure and blood pressure that has been discussed in the experimental toxicology (animal experiments) literature. That finding has important implications for the benefits of reducing lead in drinking water because high blood pressure, in turn, is linked to a variety of cardiovascular diseases.

Other recent human studies show deleterious effects of lead exposure upon fetal and post-natal growth and development, both' mental and physical, that can be correlated with exposure to lead in utero. These studies are discussed in the previous chapter on children's health effects. In this chapter, studies of reproduc­ tive effects on both men and women are summarized briefly, although no attempt is made to value these effects monetarily. This chapter contains four sections. Section A discusses the relationship between body lead levels and blood pressure in adult males, and includes studies of heart disease as relate~ to water hardness. Section B discusses some reproductive effects of lead exposure. Other health effects of lead, such as kidney function, immune system function, and hematological effects, are not discussed IV-2

in this document. The monetized and non-monetized benefits are summarized in Section C and some caveats and limitations of this analysis are found in Section D. A more complete discussion of the relationship between lead and blood pressure is included in the Addendum to the Air Quality

Criteria Document for Lead (U.S. EPA, 1986; appended to Volume 1). The methods for valuing monetarily the cardiovascular effects were developed in support of EPA's most recent rule reducing the amount of lead permitted in leaded gasoline. These methods are presented more fully in The Costs and Benefits of Reducing Lead

in Gasoline (U.S. EPA, 1985b).

IV.A. The Relationship between Blood Lead Levels and Blood Pressure This section analyzes the statistical relationship between blood lead and blood pressure. The first part provides a brief overview of human studies relating blood lead levels to blood pressure and from there links those changes to cardiovascular disease rates. The second part of this section discusses potential mechanisms and animal data related to lead's effect upon blood pressure. The Addendum to the Criteria Document for Lead (U.S.

EPA, 1986; p.A-1 to A-31) contains a much fuller analysis of this issue and serves as the basis for the summary contained here; it also includes a full bibliography. The third part of this section describes studies that have investigated the potential relation­ ship between cardiovascular disease rates and water hardness, and the possible role of lead in contributing to cardiovascular disease when present in water of different hardness. IV-3

While extremely important, these finding? are still the subject of some controversy. In addition, these results are limited in several ways. All of the monetized benefits are restricted to males aged 40 to 59, because lead is statistically correlated with blood pressure only in men, not women,* and because better data are available for that age range. In addi­ tion, most of these estimates cover only white males, because the existing studies have had insufficiently large samples of non­ whites. For these reasons, the cardiovascular health benefit estimates contained in this section are likely to understate significantly the adult health benefits of reducing lead in drinking water. The most important omissions are older males and black males of all ages.

IV.A.I. Epidemiological Studies of Blood Lead Levels and Hypertension Lead has long been associated with effects on blood pressure and the cardiovascular system, including a paper in the British

Medical Journal by Lorimer in 1886 that found that higher blood­ lead levels increased the risk of hypertension. Until recently, most of the studies focused only on hypertension and relatively high lead-exposure levels, and did not look for a continuous effect of lead on blood pressure. Others failed to find effects of lead on hypertension that were significant at the 95 percent

* Many fewer studies have investigated the relationship between exposure to lead and blood pressure in women. However, in the large-scale general population studies, while blood lead was positively correlated with blood pressure in women, it was not statistically significant (at the 90 percent confidence level). IV-4

confidence level, although most of them did find a positive

association. A stronger and

"more consistent pattern of results has begun to

emerge from recent investigations of the relation­

ship between lead exposures and increases in blood pressure or hypertension"

(Addendum to the Criteria Document, p. A-2)

throughout the range of measured blood-lead levels in various

clinically-defined, occupationally-exposed, or general population

groups.

IV.A.1.a. Occupational Studies

Kirkby and Gyntelberg (1985) evaluated the coronary risk

profiles of 96 heavily-exposed lead smelter workers with those of

non-occupationally exposed workers, matched for age, sex, height, weight, socioeconomic status, and alcohol and tobacco consumption.

There were no significant differences in life style habits, as

far as could be determined. Diastolic blood pressure was signifi­

cantly elevated among the lead workers, as was the percentage of

lead workers with ischemic electrocardiographic (ECG) changes and

some other factors. On the other hand, systolic blood pressure

and some other cardiovascular risk factors, e.g., angina pectoris,

were not significantly different. Overall, the authors concluded

that long-term lead workers have higher coronary risk profiles

than a comparable referent group and that these findings may

indicate a greater risk for major cardiovascular diseases, such

as myocardial infarctions or strokes. IV-5

Another study of about 50 occupationally-exposed workers

(de Kort et al., 1986) also showed blood pressure levels to be positively correlated with blood lead levels at near or below 60- 70 ug/dl, after controlling for confounding variables.

IV.A.1.b. Observational Studies

Moreau et al. (1982) found a significant relationship (p < 0.001) between blood lead levels and a continuous measure of blood pressure in 431 French male civil servants after controlling for age, body mass index, smoking, and drinking. In this study, the correlation was highest in young subjects, and decreased with age. The effect was stronger for systolic pressure than for diastolic pressure in both the de Kort and Moreau studies. The effect was statistically significant in the range of 12-30 ug/dl in the Moreau paper, although the effect was not large at that level. A more recent longitudinal study by Weiss et al. (1986) examined the blood-lead/blood-pressure relationship in 89 Boston policemen. This study also found a stronger correlation with systolic than diastolic blood pressure. Weiss' high-lead group had blood lead levels> 30 ug/dl. There are several recent general population studies, as well. Kromhout and Couland (1984) and Kromhout et al. (1985) studied 152 men, aged 57-67, drawn from the general population.

They found a significant relationship between blood lead and blood pressure. However, the statistical significance of the findings decreased or disappeared after eliminating the highest blood-lead subject and after multiple regression analyses were conducted IV-6 that included other determinants of blood pres~ure, such as age and body mass. The authors concluded that blood lead is probably a less important determinant of blood pressure than age or body mass.

"The above recent studies provide generally consis­

tent evidence of increased blood pressure being associat­

ed with elevated lead body burdens in adults, especially

as indexed by blood lead levels in various cohorts of

working men. None of the individual studies provide

definitive evidence establishing causal relationships

between lead exposure and increased blood pressure.

Nevertheless, they collectively provide considerable qualitative evidence indicative of significant associa­

tions between blood lead and blood pressure levels.

Particularly striking are the distinct dose-response

relationship seen for systolic pressure (correcting for age, body mass, etc.) by Moreau et al. and the

findings of significant associations between blood lead and systolic pressure after extensive and conserva­ tive statistical analyses by Weiss et al. However, estimates of quantitative relationships between blood lead levels and blood pressure increases derived from

such study results are subject to much uncertainty, given

the relatively small sample sizes and limited population

groups studied. Two larger-scale recent studiErs-·of general

population groups, reviewed next, provide better bases

for estimation of quantitative blood-lead blood-pressure

relationships." (Addendum, p. A-10) IV-7

IV.A.l.c. Population Studies

Pocock et al. (1984) evaluated the relationships between blood lead level, hypertension and renal function in a clin~cal study of 7,735 middle-aged (aged 40-49) British men. (This study, conducted by the British government is also called the British

Regional Heart Study.) In that article, the authors interpreted their findings as being suggestive of increased hypertension at elevated blood-lead levels (> 37 ug/dl), but not at the lower levels found typically in British men. However, more recent multiple regression analyses, adjusted for variation due to site, reported in Pocock et al. (1985) for the same data indicate highly statistically significant associations between both systolic

(p = 0.003) and diastolic (p < 0.001) blood pressure and blood lead levels. Noting the small magnitude of the observed association and the difficulty in adjusting for all potentially relevant confounders, Pocock et al. (1985) cautioned against prematurely concluding that there is a causal relationship between body lead burden and blood pressure.

Several other studies of this relationship, discussed below, have been published using data from the Second National Health

Assessment and Nutritional Evaluation Survey (NHANES II),* which provide careful blood lead and blood pressure measurements on a

* The NHANES II was a 10,000 person representative sample of the U.S. non-institutionalized population, aged 6 months to 74 years. The survey was conducted by the (U.S.) National Center for Health Statistics (NCHS) ·over a four-year period (1976-1980). The data base is available from NCHS and analyses of the lead­ related data from it have been published before (e.g., Annest et al., 1982 and 1983; Mahaffey et al., 1982a and 1982b~ Pirkle and Annest, 1984). IV-8

large-scale sample representative of the U.S. population and considerable information on a wide variety of potentially confound­

ing variables, as well. The studies using the NHANES II data

avoided the problems of selection bias, healthy worker effect, work place exposures to other toxic agents, and appropriate choice

of controls, that had complicated or confounded many of the other studies (cf, Addendum to the Criteria Document, 1986; p. A-12ff).

Simple correlation analyses reported by Harlan et al. (1985) demonstrated statistically significant linear associations (p < 0.001) between blood lead concentrations and blood pressure (both diastolic and systolic) among males and females, aged 12-74 years. Controlling for many potentially confounding factors, multiple regression analyses showed the blood-lead/blood-pressure

relationships remained significant for males but not for females.

Pirkle et al. (1985) conducted additional analyses on the

NHANES II data. That study focused on white males, aged 40-59,

to avoid the effects of collinearity with age, and because of less

extensive NHANES II data being available for non-whites. In the

subgroup studied, Pirkle et al. found significant associations

between blood lead levels and blood pressure both in basic models

and after including all the known factors previously established

as being correlated with blood pressure. The relationship also

held, with little change in the coefficient, when tested against

every dietary and serological variable measured in the NHANES II. "No evident threshold was found below which blood

lead level was not significantly related to blood pres­

sure across a range of 7 to 34 ug/dl. The dose-response IV-9

relationships characterized by Pirkle indicate that

large initial increments in blood pressure occur at

relatively low blood-lead levels, followed by a leveling

off of blood pressure increments at higher blood-lead

levels. Pirkle et al. also found lead to be a signifi­

cant predictor of diastolic blood pressure greater than

or equal to 90 millimeters of mercury (mm Hg),* the

criterion blood pressure level now commonly employed in

the United States to define hypertension. Additional

analyses were performed by Pirkle et al. to estimate the

likely public health implications of the Pirkle findings

concerning the blood-lead/blood-pressure relationship.

Changes in blood pressure that might result from a speci­

fied change in blood lead levels were first estimated. Then coefficients from the Pooling Project and Framingham

studies (Pooling Project Research Group, 1978; and McGee

and Gordon, 1976, respectively)** of cardiovascular disease were used as bases: (1) to estimate the risk for incidence

of serious cardiovascular events (myocardial infarction, stroke, or death} as a consequence of lead-induced blood

pressure increases and (2) to predict the change in the number of serious outcomes as the result of a 37 percent

decrease in blood lead levels for adult white males (aged

40-59 years) observed during the course of the NHANES II

0 survey ( 1976-1980) .• ' (Addendum, p. A-13}.

* Millimeters of mercury is the standard measure of blood pressure.

** These are discussed more fully in Section IV.A.4., below. IV-10

Schwartz (1985) expanded the Pirkle et al. analysis to include all men over 20, and examined the 20-44 and 45-74 age_groups separately. In all three groups, lead was significant both in basic models and when ~ested against the much-larger variable list. Schwartz also showed that lead was significant even with the inclusion of a linear time-trend term.

"Questions have been raised by Gartside (1985) and

Du Pont (1986) regarding the robustness of the findings

derived from the analyses of the NHANES II data as to

whether certain time trends in the NHANES II data set

may have contributed to (or account for) the reported

blood-lead/blood-pressure relationships •••

However, neither the Gartside nor the Du Pont analyses

adjusted for all of the variables that were selected for stepwise inclusion in the Harlan et al. (1985) and

Pirkle et al. (1985) published studies." (Addendum, p. A-14)

Reanalyses of the NHANES II data by Schwartz (1986) showed that lead remained a significant factor even if a linear time trend was included in the regression, and also remained significant in regressions that adjusted for all of the sites visited in NHANES II, which controls for both time (the sites were visited sequenti­ ally) and geographical variation. Schwartz also showed that the relationship held in all adult men, in men under 45, and in men over 45, as well as in the age group of Pirkle et al. These reanalyses were reviewed and accepted by EPA's external Science

Advisory Board. IV-11

"In order to more defin~tively assess the robust­ ness of the Harlan et al. {1985) findings and, aiso,

to evaluate possible time-trend effects confounded by

variations in sampling sites, Landis and Flegal {1986)

carried out further analyses for NHANES II males, aged

12-74, using a randomization model-based approach to test the statistical significance of the partial cor­

relation between blood lead and diastolic blood pres­ sure, adjusting for age, body mass index, and the 64 NHANES II s~mpling sites. The resulting analyses con­

firm that the significant association between blood lead

and blood pressure cannot be dismissed as spurious due to concurrent secular trends in the two variables over

the NHANES study period." {Addendum, p. A-14) In addition, EPA has conducted a series of additional reanaly­ ses of the NHANES II data to address the issue of "site" more definitively. "These unpublished analyses* confirm that the regres­ sion coefficients remain significant for both systolic

and diastolic blood pressure when site is included a~ a

variable in multiple regression analyses." (Addendum, p. A-15)

* Available in the Central Docket Section of EPA. Docket number, ECAO-CD-81-2; documents numbered IIA.F.60, IIA.C.5, II.A.C~9, and IIA.C.11. IV-12

"Overall, the analyses of data from the two large­ scale general population studies (British Regional Heart

Study and U.S. NHANES II study), conducted both in this country and in Great Britain, collectively provide highly convincing evidence demonstrating small but statistically significant associations between blood lead levels and increased blood pressure in adult men.

The strongest associations appear to exist for males aged 40-59 and for systolic pressure somewhat more than for diastolic. Virtually all of the analyses revealed positive associations for the 40-59 age group, which remain or become significant (at p < 0.05) when adjust­ ments are made for geographic site. Furthermore, the results of these large-scale studies are consistent with similar findings of statistically significant associations between blood lead levels and blood pressure increases as derived from other recent smaller-scale studies dis­ cussed earlier, which also mainly found stronger associa­ tions for systolic pressure than for diastolic. None of the observational studies in and of themselves can be stated as definitively establishing causal linkages between lead exposure and increased blood pressure or hypertension. However, the plausibility of the observed associations reflecting causal relationships between lead exposure and blood pressure increases is supported by: (1) the consistency of the significant associa- tions that have now been found by numerous independent IV-13 investigators for a variety of study populations; and

(2) by extensive toxicological data (see below) which clearly demonstrate increases in blood pressure for animal models under well-controlled experimental con­ ditions. The precise mechanisms underlying the relqtion­ ships between lead exposure and increased blood pressure, however, appear to be complex, and mathematical models describing the relationships still remain to be more definitively characterized. At present, log blood-lead/ blood-pressure (log PbB-BP) models appear to fit best the available data, but linear relationships between blood lead and blood pressure cannot be ru~ed out at this time. The most appropriate coefficients charac­ terizing blood-lead/blood-pressure relationships also remain to be more precisely determined, although those reported by Landis and Flegal (1986) and those obtained by analyses adjusting for site appear to be the currently best available and most reasonable estimates of the likely strength of the association (i.e., generally in the range of 2.0-5.0 for log PbB versus systolic and 1.4 to 2.7 for log Pbij versus diastolic blood pressure)." "Blood lead levels that may be associated with increased blood pressure also remain to be more clearly defined. However, the collective evidence from the above studies points toward moderately elevated blood­ lead levels (~ 30 ug/dl) as being associated most clearly with blood pressure increases, but certain evidence IV-14

(e.g., the NHANES II data analyses and the Moreau et al. study results) also indicates significant (and apparently stronger) relationships between blood pressure elevations and still lower blood lead levels that range, possibly, to as low as 7 ug/dl. This may be supported by several animal studies, discussed below, that also find hypertension most consistently related to relatively

low exposure levels but over relatively long exposure periods." "The quantification of likely consequent risks for

serious cardiovascular outcomes, as attempted by Pirkle et al. (1985), also remains to be more precisely charac­ terized. The specific magnitudes of risk obtained for

serious cardiovascular outcomes in relation to lead

exposure, estimated on the basis of lead-induced blood pressure increases, depend crucially upon: the form of

the underlying relationship and size of the coefficients

estimated for blood-lead/blood-pressure associations; lead exposure levels at which significant elevations in blood pressure occur; and coefficients estimating rela­

tionships between blood pressure increases and specific, more-serious cardiovascular outcomes. As noted above,

uncertainty still exists regarding the most appropriate model and blood-lead/blood-pressure coefficients, which makes it difficult to resolve which specific coefficients

should be used in attempting to project more serious

cardiovascular outcomes. Similarly, it is difficult to IV-15

determine appropriate blood-lead levels at which any

selected coefficients might be appropriately applied in

models predicting more serious cardiovascular outcomes. Lastly, the selection of appropriate models and coeffi­

cients relating blood pressure increases to more serious

outcomes is also fraught with uncertainty •••

Further analyses of additional large-scale epidemiologic

data sets may be necessary in order to determine more

precisely quantitative relationships between blood lead

levels and blood pressure, and more serious cardiovascu­

lar outcomes as well."

"The findings discussed here, while pointing toward

a likely causal effect of lead in contributing to increased.

blood pressure, need to be placed ·in broader perspective

in relation to other factors involved in the etiology of

hypertension. The underlying causes of increased blood

pressure or "hypertension" (diastolic blood pressure

above 90 mm Hg), which occurs in as many as 25 percent of Americans, are not yet fully delineated. However,

it is very clear that many factors contribute to develop­

ment of this disease, including hereditary traits,

nutritional factors and environmental agents." (Addendum,

P• A-15 to A-18) The contribution of leadu compared to many other factors such as age, body mass, and smoking~ appears to be relatively small, but the findings have stayed robust in the face of repeate9 reanalyses and specifications of the models. IV-16

IV.A.2. Mechanisms Potentially Underlying Lead-Induced Hypertension Effects "This section [briefly summarizes] .plausible bio­

chemical-physiological mechanisms by which lead poten­

tially influences the cardiovascular system to induce

increased blood pressure, followed by [a short discussion]

of experimental evidence concerning the contribution of

lead exposure to the development of hypertension.

"Blood pressure is determined by the interaction

of two factors: cardiac output and total peripheral

resistance. An elevation of either or both results in

an increase in blood pressure. A subsequent defect in a

critical regulatory function (e.g., renal excretory

function) may influence central nervous system regulation

of blood pressure, leading to a permanent alteration in

vascular smooth muscle tone which sustains blood pressure

elevation. The primary defect in the pathophysiology of hypertension is thought to be due to alteration in cal­

cium binding to plasma membranes of cells; this change

in calcium handling may in turn be dependent upon an

alteration in sodium permeability of the membrane (e.g.,

Hilton, 1986). This change affects several pathways

capable of elevating pressure: one is a direct altera­

tion of the sensitivity of vascular smooth muscle to

vasoactive stimuli; another is indirect, via alteration

of neuroendocrine input to vascular smooth muscle (includ­

ing changes in renin secretion rate)." (Addendum, p. A-18) IV-17

IV.A.2.a. Role of Disturbances in Ion Transport by Plasma Membranes "Many stimuli activate target cells in the mammalian

body via changes in ion permeabilities of the plasma

membrane, primarily for sodium, potassium, and calcium

ions; the change in calcium ion concentration is the·

primary intracellular signal controlling muscle contrac­

tions, hormone secretion, and other diverse activities •••

For calcium, there is a membrane potential-dependent

sodium/calcium exchange pump which extrudes one calcium

ion in exchange for three sodium ions. In addition,

there are calcium ATPase pumps located at cell membranes

and at intracellular membrane storage sites (endoplasmic

reticulum and mitochondria) ••• The ion interacts with several calcium-binding proteins which, in turn, activate

cell contractile or secretory processes.

"It has been postulated that sodium pump inhibition by some endogenous factor (thought to be a hormone)

would be ultimately causatory for development of both

essential and volume-expanded hypertension by affecting

vascular tone or resistance.

"If lead exposure could be shown to affect sodium

transport (which, then, indirectly alters vascular resistanc~) or to directly affect vascular resistance

(by changing calcium ion permeability or transport), it could contribute to the development of hypertension ••• IV-18

Abundant experimental evidence ••• indicates that

lead affects both; that is, lead inhibits cell membrane­

bound sodium-potassium-ATPase as well as interferes with

normal processes of calcium transport across membranes

of various tissue types ••• Lead acts to alter sodium

balance and calcium-activated cell activities of

vascular smooth muscle. Changes in either or both of

these could be expected to produce changes in blood

pressure regulation." (Addendum, p. A-18 to A-19)

IV.A.2.b. Role of Renin-Angiotensin in Control of Blood Pressure and Fluid Balance

One major endogenous factor regulating total peripheral resistance of the vascular smooth muscle is angiotensin II (AII), a small peptide generated in plasma via the action of a renal hormone, renin. "The renin-angiotensin system has a major influence on regulation of blood pressure, [both directly and

indirectly. It directly affects vascular smooth muscles to increase vasoconstriction and it indirectly increases total peripheral resistance by affecting the discharge

rate of sympathetic neurons.] For this reason, investi­

gators interested in hypertension have studied the

system in detail. Because renal disease may be an important initiating event in subsequent development of hypertension and because lead is an important renal

toxicant, some investigative reports of patients with lead intoxication have evaluated blood pressure changes IV-19

·and changes in the renin-angiotensin system." (Addendum,

P• A-20 to A-21)

However, the results of these studies have been contradictory. The few human studies have shown depressed, increased and unaltered renin activity in lead intoxicated men. The studies may not be comparable because some used men with chronic sub-clinical lead exposure as compared to chronic heavy lead exposure, and some subjects had exposure which would be considered "normal".

In addition, there have been animal and experimental studies, investigating the cardiovascular effects of both acute and chronic exposure to lead.

"Lead injected intravenously in dogs and rats, at

doses as low as 0.1 mg/kg (whole blood lead< 5 ug/dl

and renal lead of 1.2 ug/g) produced over the next

several hours significant increases in plasma renin activity (PRA) and in excretion of sodium, other cations,

and water (Mouw et al., 1978) ••• The increased sodium excretion could be attributed to decreased sodium reabsorp­ tion. The mechanism of lead's action on tubular reabsorp­ tion was not determined, ••• nor was the mechanism by which

lead increased renin secretion. "In a subsequent study, Goldman et al. (1981) found

that the rise in PRA after acute lead injection was not due

to increased renin secretion in six of nine dogs; rather there was elimination of hepatic renin clearance, without

evidence for other interference in liver function. In the remaining three dogs, renin secretion increased; this was thought to be due to lead activation of n·orrnal mechanisms IV-20 for renin secretion, although none of the classic pathways for influencing renin secretion were altered.

The authors postulated that lead might produce altera­ tions in cytosolic calcium concentration in renin­ secreting cells ••• The authors also postulate that there may be multiple actions of lead on the renin-angio­ tensin system which may help explain confusion about the ability of lead to cause hypertension. At certain exposure conditions, there could be elevated PRA without simultaneous inhibition of angiotensin-converting enzyme, thereby contributing to hypertension, while higher doses or longer exposure might inhibit the converting enzyme and thereby cause loss of hypertension •••

"The literature of experimental findings of lead­ induced changes in the renin-angiotensin system and blood pressure in animals is complicated by apparently inconsistent results when comparing one study to another.

All studies report changes in the renin-angiotensin system, yet some studies fail to find an effect on blood pressure and others do report hypertension. Doses and exposure periods employed vary widely, but in general, hypertension is observed most consistently with relative­ ly low doses over relatively long exposure periods •••

"Perry and Erlanger (1978) found that chronically feeding rats either cadmium or lead at doses of 0.1,

1.0, or 5.0 parts per million (ppm) produces statistically significant increases in systolic blood pressure ••• IV-21

The implications for human populations exposed to very low doses of these metals were pointed out. Victery et al. (1982a) reinvestigated the question, using lead doses of 100 and 500 ppm administered in the drinking water to rats beginning while the animals were in utero and continuing through six months of age. At 3-1/2 months of age, the male rats drinking 100 ppm of lead first demonstrated a statistically significant increase in systolic blood pressure; this difference persisted for the remainder of the experiment. Animals drinking

500 ppm had lower pressures, which were not significantly different from controls. Female rats drinking 100 ppm did not demonstrate pressure changes. At termination of the experiment PRA was significantly decreased by 100 ppm lead exposure, but not 500 ppm ••• There was a dose-dependent decrease in the ratio of [angiotensin II to plasma renin activity] for lead-exposed rats. Renal renin was depressed in lead-exposed animals. The hyper­ tension observed in these animals was not secondary to overt renal disease (as opposed to an effect on renal cell metabolism), as evidenced by a lack of changes in renal histology and plasma creatinine.

"Vict~ry et al. (1983) examined changes in the renin­ angiotensin system of rats exposed to lead doses of 5, 25, 100, or 500 ppm during gestation until one month of age. All had elevated plasma renin activity, while those at 100 and 500 ppm also had increased renal renin IV-22

concentration. Lead-exposed animals ••• secreted less renin than control animals. It appears that lead

has two chronic effects on renin secretion, one inhibi­ tory and one stimulatory; the magnitude of effect on PRA

reflects the dose and timing of the lead exposure as

well as the physiological state of the animal. "In another study, Victery at al. (1982b) reported

that rats fed 5 or 25 ppm lead for five months (blood lead of 5.6 and 18.2 ug/dl, respectively) did not develop

hypertension but at 25 ppm had significantly decreased

PRA. Both groups of animals had a decrease in the.AI!

to PRA ratio. Thus, lead exposure at leve,ls generally present in the human population caused observable'effects

in renin synthesis." (Addendum, p. A-23 to A-25)

IV.A.2.c. Effects of Lead on Vascular Reactivity

"Piccinini et al. (1977) and Favalli et al. (1977) studied the effects of lead on calcium exchanges in the

isolated rat tail artery; lead in concentrations of up to 15 umol in vitro produced contractions which required

the presence of calcium in the perfusion solution. Therefore, calcium influx was not affected by lead •••

Tissue calcium content was increased ••• "Tail arteries obtained from the hypertensive rats

in the study performed .by Victery et al. showed an

increased maximal contractile force when tested in vitro

with the alpha-adrenergic agents norepinephrine and IV-23

methoxami.ne (Webb et al., 1981). This finding is apparently

related to an increase in the intracellular pool of activator

calcium in the smooth muscle cells in the artery. This

change may also be responsible for decreased relaxation

of the muscle after induced contractions.

"In vivo tests of cardiovascular reactivity in rats

exposed to 50 ppm lead (blood lead 38.4 + ug/dl) for 160

days were performed by Iannaccone et al. {1981). [This

study showed] significant increases in systolic and diastolic

pressure [related to lead exposure, as well as significant increases in the blood pressure response to noradrenalin].

The data suggest that the lead-related increase in arterial

pressure is due at least in part to greater sympathetic

tone, with the metal affecting neural control of blood

pressure." (Addendum, p. A-26 to A-27)

IV.A.2.d. Effects of Lead on Cardiac Muscle

Lead has been hypothesized to contribute to cardiomyopathy and to have cardiotoxic properties.

"Kopp et al. {1978) developed an in vitro system for

monitoring the cardiac electrical conduction system {electrocardiogram or ECG) and systolic tension, and

demonstrated that in vitro lead {3 x 10-2 mM) or cadmium

{3 x 10-2 mM) depressed systolic tension and prolonged the

P-R interval of the ECG ••• [In a later study,] hearts

obtained from rats exposed to low levels of cadmium and/or

lead {5 ppm) for 20 months were found to have ••• changes in the heart's electrical conduction system {Kopp et al., 1980) with significant prolongation of the P-R interval ••• IV-24

"Williams et al. (1983) suggested that much of the negative effect of lead on cardiac tissue and ECG abnor­ malities can be related to lead's interference with calcium ion availability and/or membrane translocation.

In addition, even those lead exposure-related effects that appear to occur through autonomic nerves may be understood in terms of effects on calcium ion, which is required for neurotransmitter release •••

"Prentice and Kopp (1985) examined functional and metabolic responses of the perfused rat heart produced by lead with varying calcium concentrations in the per­ fusate. Lead altered the spontaneous contractile activity, spontaneous electrical properties and metabo­ lism of the heart tissue. The exact mechanisms were not completely resolved but did involve disturbances in cellu­ lar calcium metabolism, although not by any single mechanistic model •••

"In addition, hearts perfused with 30 uM lead had reduced coronary blood flow, presumably by lead acting to directly constrict the vascular smooth muscle or by interference with the local metabolic stimuli for vaso­ dilatation. Increases in perfusate calcium concentration partially reversed this effect, although at the highest calcium levels (5.0 mM) coronary blood flow was again reduced. These authors concluded that their present findings were consistent with those of others which IV-25

showed increased vascular reactivity and that the

chronic lead exposure-related changes in blood pressure

may be related to localized actions of lead on vascular

beds and arterial "Smooth muscle." (Addendum, p. A-27 to

A-29)

IV.A.2.e. Summary of Lead-Related Effects on the Cardiovascular System

"Blood pressure is regulated and affected by many

interactive forces and control systems; some of these

have. been shown to be affected by lead exposure. Under­

standing of the effects of lead on each system is still preliminary, but sufficient evidence indicates that

changes which occur in the presence of lead can promote

the development of hypertension ••• Although the

exact mechanisms involved in lead-induced changes in renin secretion rate have not been examined, it is

likely that lead could be affecting the cystosolic free calcium ion of [some] cells ••• After lead enters

[these] cells, lead could enhance or block calcium exit

via sodium/calcium exchange pumps, or increase or decrease the intracellular sequestration of calcium in

storage compartments ••• "The changes in vascular reactivity which have been reported in animals chronically exposed to lead are

probably the key finding which can lead to an under­

standing of how lead can contribute to the development IV-26

of hypertension. The vascular smooth muscle changes

are necessary and sufficient in themselves to account for the increase in blood pressure and the fact that

these changes are observed in animals exposed to rela­

tively low lead levels makes it increasingly important

to evaluate these findings in additional experimental

studies. There may be additional changes in the entire

sympathetic neural control of vascular tone which acts

to amplify the contractile response to any endogenous

vasoconstrictor substance." (Addendum, p. A-29 to A-30)

IV.A.3. Cardiovascular Disease Rates and Water Hardness

Since Kobayashi's study (1957) was highlighted by Schroeder

{1958), studies in several countries have evaluated the relation­ ship between drinking water characteristics and death from hypertensive, ischemic or arteriosclerotic heart diseases, and in particular the relationship between cardiovascular disease (CVD) rates and water hardness.* Generally, the studies covering large geographic regions have shown a significant inverse rela­ tionship between the hardness of local drinking water and local CVD rates. However, not all the studies, most commonly not the studies involving smaller geographic regions, evidence the same consistency of effect.

* Water hardness is determined by the relative amount of dissolved solids, primarily calcium and magnesium, in it. It is expressed as the equivalent amount of calcium carbonate (CaC03) that could be formed from the calcium and magnesium in solution. For a fuller discussion of this issue, see Chapters II and V of this document. Generally, water< 60 mg/1 as CaCO3 is considered 'soft'. Most studies investigating CVD and water hardness have focused upon systems with very soft water, i.e.,< 40 mg/1 as CaC03. - IV-27

The U.S. National Research Council, of the National Academy of Sciences, convened a panel of experts in the late 1970s to review this issue. The panel concluded that the findings we~e

"strikingly" equivocal and called for additional research (Angina,

1979}. Since then, many additional reviews have attempted to resolve the inconsistencies in the data; none has succeeded. Two noteworthy reviews (Comstock, 1979; Comstock, 1985) apply rigorous statistical and epidemiologic tests to the literature on the relationship between water hardness and CVD. This section contains a very brief survey of the nature of these studies, plus a suggest­ ed explanation for the inconsistent findings.

IV.A.3.a. Studies of Cardiovascular Disease and Water Hardness

The largest category of studies is the intra-national studies, where various population units within several different countries were examined. Generally, these studies have found a statistically significant inverse relationship between CVD and water hardness, and account for the phrase: 'soft water, hard arteries'.

A second category of studies -- geographic units comprising states or provinces -- in general show a similar but weaker negative association between CVD mortality and water hardness.

Studies of selected communities that are not in the same geographic unit and that have different water hardnesses have been least likely to show consistent findings linking CVD and water parameters; the inconsistencies of findings have included associations with water hardness that were in opposite directions for the two sexes or for different categories of cardiovascular illnesses, as well as statistically insignificant associations. IV-28

Fewer of the studies conducted in the past decade have found the significant negative associations that were found in some of the earI 1 ier• papers.

Interest in identifying a 'water faetor' that influences heart disease has continued because CVD rates vary significantly by geographic regions {Sauer, 1980). Water parameters also tend to portray regional patterns and, therefore, could offer a tenable explanation for the geographic variation in mortality. Further, water treatment costs are lower than medical treatment costs.* At the water system level, reducing the presence of a harmful sub­ stance or correcting the deficiency of a protective one is both relatively easily done and effective in reaching large numbers of people.

There are many plausible mechanisms for recognizing an effect from the constituents and parameters of water upon heart disease.

The presence or selective absence of trace elements are the most likely sources of an effect on health related to corrosive water.

Or soft water may initiate heart disease, or soft water parameters added onto other factors that predispose a person to heart disease may be sufficient to push him/her over the threshold to symptomatic illness, or it may exacerbate a pre-existing condition, either recognizable or asymptomatic. Soft water may eontain harmful contaminants or hard water may contain protective elements laeking in soft water. Most probably, some combination of these scenarios accounts for the association.

* For instance, treatment to increase water hardness costs $1-2 per person per year as opposed to orie heart attack, whieh costs $65,000 per case. IV-29

Generally, there are three broad classes of mechanisms by which water can be postulated to affect CVD in either a causative or detrimental manner.

1) Soft water is deficient in some major cations that are protec­

tive to the cardiovascular system and that are abundant in

hard water. The most likely of these are calcium and magne­

sium, which are the major components of hardness in water.

Some epidemiological studies (e.g., Abu-Zeid, 1979: though

not all, e.g., Miller et al., 1985) have shown magnesium to

be a strong factor in cardiovascular health.

2) Protection of the cardiovascular system by specific minor

cations or trace metals that are more abundant in hard water than soft water. Of these, lithium, chromium, manganese,

selenium, vanadium, and strontium are possible contributors. All, with the possible exception of vanadium, are essential

micronutrients and could have plausible involvement with CVD

rates.

3) Soft water leaches toxic metals from the distribution system. Studies relating water hardness to CVD have considered

especially three potential toxins: cadmium, copper, and

lead. Copper is perceived as the least likely of the three, and strong evidence exists linking both lead and cadmium to

hypertens~on. All three metals are corrosion by-products.

(It is interesting, in this light, to note that water is a more significant source of lead in Britain than the u.s. and the hard-water/CVD association appears more robust there.) IV-30

Overall, the two constituents most seriously considered in current analyses are the protective role of magnesium (less common in soft water than in hard water) and a detrimental effect from increased sodium (which is often added in water softening units).*

IV.A.3.b. Lead, Soft Water and Cardiovascular Disease While the geographic variation in CVD probably relates to several factors and the interrelationships between them, the new studies relating blood lead levels to blood pressure could help resolve some of the inconsistent correlations between cardiovascular disease and soft drinking water. Lead occurs in drinking water as a corrosion by-product, that is, as a result of the action of corrosive water upon the materials (pipes and particularly solder) of the distribution and residential plumbing systems.** Hardness/ softness of water is one measure of a water's potential corrosivity, but because water chemistry is very complicated, no single measure is an adequate predictor of a specific water's actual corrosivity or of the presence of corrosion by-products, including lead (cf, for instance, Patterson, 1981). Relatively few of the studies investigating the relationship between water hardness and CVD rates have considered the presence

* It has been suggested that the relatively recent increase in sodium intake in populations in hard water areas may be negating the advantages once associated with those hard water areas · (Comstock, 1985). ** For a fuller discussion of these studies, see Section II.A., above, or Section V.A., below. IV-31 of trace metals at all (e.g., Sharrett et al., 1984) or the water sampling procedures have not addressed the occurrence phenomena of corrosion by-products.* Often, the water data represent lead levels in the distribution system (e.g., Shaper et al., 1980; or

Craun and McCabe, 1975, which uses data from the 1969 Community

System Survey, presented in McCabe et al., 1970); or tap water typical of the distribution system· (i.e., fully flushed); or tap water averaged with distribution system water (e.g., Greathouse and Craun, 1978; or Greathouse and Osborne, 1980; which use data from the First National Health and Nutrition Augmentation Survey, in which one random daytime tap sample was averaged with 12 monthly samples from the water supply).

In the few studies that did consider exposure to corrosion by-products, other problems have arisen. In Calabrese and Tuthill

(1978), for instance, the lead levels in the two communities studied were not statistically significantly different. Other studies have not withstood rigorous statistical testing (e.g., Hewitt and Neri, 1980). Considering the data linking blood pressure (and CVD) to exposure to lead, the studies correlating CVD with water hardness may be imperfectly measuring the relationship between soft drinking water and the presence of lead (and/or cadmium). In other words, the studies that have found a relationship between soft water and heart disease may actually have evaluated systems with higher-than-

* That is, contamination of drinking water that occurs due to the corrosive action of the water upon the materials of the plumbing systems. Several of the trace constitu~nts of concern, including lead, copper and cadmium, are corrosion by-products. This issue is discussed in Chapter II of this document. IV-32 average lead levels, resulting from softer water leaching more lead from the plumbing. However, while softer water is more likely to contain higher lead (or cadmium) levels, the factors related to lead contamination are very complicated. Soft water alone is not predictor of lead levels, and therefore, not all systems with soft water have higher metals. This may account for some of the inconsistent findings.

Another factor is that the major elements of hard water are magnesium and calcium which can be easily absorbed in the.small

intestine and therefore may compete with lead for some common transport system or metabolic interactions, as explained by

Conrad and Barton (1978) and Mahaffey and Rader (1980). The relationship between calcium intake and lead absorption is not yet clear. Some early studies (e.g., Aub, 1935) indicated that

increased calcium intake could reduce blood lead levels; other studies have shown more complicated relationships between calcium

intake and lead absorption and retention, including lead's effect upon calcium metabolism (Six and Goyer, 1970 and 1972; Quarterman. et al., 1978; also cf the discussions on lead and calcium in both the Criteria Document, 1986, p. 10-44 to 10-48, and its Addendum, p. A-18ff).

IV.A.4 Benefits of Reduced Cardiovascular Disease: Reductions in Hypertension and Related Morbidity and Mortality

On the basis of the data discussed above on the assoc-iat-i0n between lead body burden and increased blood pre~sure in adult males, this analysis assumes that reducing lead in drinking water

could reduce blood lead levels, which in turn could reduce blood IV-33 pressure and the number of individuals with hypertension. A reduction in hypertension would have direct benefits from reduced· medical treatment expenditures. More important, however, are the related benefits in the form of reduced cardiovascu+ar disease associated with elevated blood pressure. This section describes the methods used to estimate the benefits associated with lowering blood pressure, including estimating the reductions in morbidity and mortality.

Estimating the reduction in hypertension and cardiovascular disease requires several steps. The first is to estimate the impact of the reduction of lead levels in drinking water on levels of lead in adults' blood. For that, the occurrence data and water-lead to blood-lead equations described in Chapter II were combined with the regression analyses of the NHANES II data discussed previously. In each case, the blood lead levels in the

NHANES II data were first adjusted to reflect reductions in environmental lead contamination that have occurred since the time of the survey. To calculate the cardiovascular benefits, logistic regression equations were used to predict how reducing exposure to lead in drinking water could affect the number of hypertensives in the U.S. population. These estimates cover only males aged 40 to 59, because the effect of lead on blood pressure appears to be stronger for men and because the correlation between blood pressure and age is much smaller in this age range, reducing the potential for confounding due to the correlation between blood lead and age. The estimates of reduced cases of cardiovascular and IV-34 cerebrovascular disease rely upon (1) site-adjusted coefficients from analyses of the NHANES II data relating blood lead levels to increases in blood pressure* and (2) coefficients relating blood pressure increases to more serious cardiovascular disease outcomes, based on data from the Framingham Study and Pooling Project (1976), discussed below. The estimates of myocardial infarctions, strokes and deaths are further restricted to white men, aged 40-59, because the

NHANES data contain insufficient observations on non-whites to evaluate the form of the relationship among non-whites. This severely limits this analysis because both blood pressure and blood lead levels are higher among non-whites than whites.**

The fact that other sources of lead, especially gasoline, would slowly decline even without new EPA drinking water standards created a slight complication. Because gasoline lead levels fall over time as unleaded gasoline replaces leaded, the difference in blood lead levels resulting from this rule will change over time. The estimates in this report account for both the reduction

* The specific coefficients and the basis for their derivation are described in the Addendum to the Criteria Document, 1986, which is included in Volume 1 of that publication. The issue of site adjustment is summarized briefly on p. IV-lOff of this document.

** In other EPA analyses, e.g., Methodology for Valuing Health Risks of Ambient Lead Exposure (US-EPA, 1986a), less--conse-rva­ tive assumptions have been made -- i.e., assuming that the same cardiovascular risks apply to black men as to white men of the same age (cf., pages 5-5 ff). Extending this current analysis to black men, even using the same coefficients and assumptions as for white men, would raise the estimated benefits. IV-35

in blood pressure over the past decade and reductions in some other sources of lead. This model served EPA also in its analy­ tical efforts supporting the most recent phasedown in the amount of lead permitted in leaded gasoline; it is discussed more fully

in The Costs and Benefits of Reducing Lead in Gasoline (US-EPA,

1985b). The coefficients were adjusted for site-variation as discussed in the Addendum to the Criteria Document (US-EPA, 1986; p. A-13ff).

IV.A.4.a. Hypertension

Based upon the studies relating lead exposure to blood pressure, discussed above, estimating the change in the number of cases of hypertension was straightforward: the logistic regres­ sion coefflcients from Pirkle et al. (1985) adjusted for site were applied to the NHANES II data to predict the changes in the numbers hypertensives as a result of reducing the Maximum Contaminant

Level (MCL) for lead in drinking water. In this analysis, EPA estimates that there would be 130,000 fewer cases of hypertension among males aged 40-59 in sample year 1988 as a result of this

rule. The change due to this proposed regulation was calculated

by subtracting the number at the new lead level from the number at the original lead level. This estimate covers only males aged

40 to 59, but includes non-whites as well as whites~

IV.A.4.b. Myocardial Infarctions, Strokes, and Deaths

Estimating the impact of reduced blood pressure on morbidity

and mortality required several additional steps. Using the NHANES IV-36

II data and the site-adjusted regression coefficients described earlier, we simulated the changes in individual blood-pressure levels due to reductions in lead from drinking water. Coefficients from two large studies of cardiovascular disease were then used to estimate changes in the numbers of first-time myocardial infarc­ tions, first-time strokes, and deaths from all causes.

The relationships between blood .pressure and-cardiovascular diseases have been established by several large, long-term epidemio­ logical studies. The classic study, which was important in establishing cholesterol as a major factor in the risk of heart disease, was the Framingham study (McGee et al., 1976). Extensive analyses of these data have yielded estimates of cardiovascular risks associated with several variables, including blood pressure.

Figure IV-1 shows the age-adju~ted rates of death and heart attacks as functions of blood pressure from that study. In the 1970s, the National Institutes of Health funded the

Pooling Project_(The Pooling Project Research Group, 1978), which combined the Framingham data with data from five other long-term studies to improve the accuracy of the risk coefficients for heart attacks. The Pooling Project tested the Framingham coefficients against the other study results and found that their predictive power was good. It then analyzed the first occurrence of myocar­ dial infarctions in white men who entered the studies at ages 40 to 59 and who were followed for at least 10 years. The estimates of the numbers of first-time myocardial infarctions in this study employ the Pooling Project's coefficients. IV-37

FIGURE IV-1

Adjusted Rates of Death and Heart Attacks versus Blood Pressure: Framingham.Data

ANNUAL ~ OF IFAn1 BY DIAS'roLIC BIDCD PRE.5SURE

Males 45-74 (age adjusted rate)

250 - m C 225 - 1-1g CD C. C - C 200 C. C .-1 175 - 1-1 !. 150 -

125 - i----4

100 - _,,,_...

(. 70 70- 75- 80- 8S- 90- 95- 100- 105- 110+ 74 79 84 89 94 99 104 109

ANNUAL INCID.EOCE OF MroCARDIAL INFARCl'IOO BY DIAS'IDLIC BLCCO PRESSURE

Males 45-74 (age adjusted rate)

Ill · 100 C - 0 s..Cl) CD 90 C. - C C C. 80 - C .-1 1-1 G) 70 - c._ Ill C, C 60 - I G) ....'O C, 50 - ....C .-1 I'll :, 40 - C C ,c -- C 70 70- 75- 80- 85- 90- 95- 100- 105- 110+ 74 79 84 89 94 99 104 109 IV-38

In addition to estimating the risk of heart attacks, the

Framingham study estimated regression equations for the risks of stroke and death as functions of blood pressure and other vari­ ables. Because the Pooling Project did not include those end­ points, the Framingham study coefficients are used in this analysis. As with heart attacks, the estimates for strokes cover only first-time events; thus, the estimates for strokes and myocardial infarctions are biased downwards because they exclude second and subsequent heart attacks and strokes associated with elevated blood pressure. The regression equation for deaths covers all CVD-related causes of death; it includes deaths not just from myocardial infarctions and strokes, but also from other causes associated with blood pressure (e.g., heart diseases other than myocardial infarctions). Levy et al. (1984) recently tested the Framingham study regression coefficients to see how well they explained the observed decrease in cardiovascular mortality in the United States from 1970 to 1980. They found that the coefficients, when coupled with changes in blood pressure and other cardiovascular risk factors over that same period, were able to explain about 80 percent of the drop in cardiovascular mortality.

There is also clinical evidence showing that increased (or decreased) blood pressure can be associated with cardiovascular events and mortality rates. For instance, the Hypertension

Detection and Follow-up Program (New England Journal of Medicine,

1982) found that intervention resulting in about a 5 mm Hg change in diastolic blood pressure produced a 20 percent reduction in IV-39 overall mortality. The Australian National Trial on mild hyper­ tension also found reductions in morb1dity and mortality resulted from lowered blood pressure (Lancet, 1980). The Multiple Risk

Factor Intervention Trial found that drug therapy to lower blood pressure reduced cardiovascular disease in persons with normal resting electrocardiograms (ECGs), but increased it in persons with abnormal resting ECGs (Journal of the American Medical

Association, 1982). This suggests an adverse effect of the drugs used. To produce estimates for all 40 to 59 year old white males, the individual risk of each person sampled in the NHANES II was summed and then averaged. Since the sampled individuals repre­ sent the U.S. population for their specific age-race-sex category, their average risk represents the average risk for all 40 to 59 year-old white men. Because blood lead levels have dropped since the NHANES II period, we corrected for that change and then evalua­ ted the effects of the potential MCL for lead. Again, only white men were examined because there were too few blacks in the Framingham study, and their risk might be different from whites.

The three cardiovascul?r-risk regression equations all predict risk over the next 10 years, given current blood pressure, age, and other characteristics. Presumably, the risk in years 2-10 was affected by blood pressure in those years as well as by initial blood pressure. Because blood pressure levels over time in the same individual are positively c6rrelated, it is likely that the regression coefficient in part included the effect of future blood pressure levels. Lacking any data with which to estimate IV-40 the pure effect of a one-year change in blood pressure, we divided the coefficient for 10-year risk by 10. The adjusted coefficient was then used with the year-by-year predicted changes in blood pressu~e to estimate risk reductions. This procedure almost certainly overcompensates, lending a downward bias to the results, because current blood pressure is not perfectly correlated with future blood pressure.

In this analysis, we adjusted the population at risk for the increases in the u.s. population of white males aged 40 to 59. The regression from the Framingham study predicting deaths for men aged 40 to 54 was extended to 40 to 59 for data comparability and uniformity. Because the death rate actually increases with age, this also will bias the results downward. In this analysis, EPA estimates that there would be 240 fewer myocardial infarctions, 80 fewer strokes, and 240 fewer deaths among the members of the target groups in sample year 1988 as a result of the potential MCL. Extending this analysis to men of other ages and to non-whites would substantially increase these estimates.

IV.B. Lead's Effects upon Reproductive Function

At high levels, lead's adverse effects upon human reproductive function have been known for over 100 years.* In 1860, for instance, Paul published findings that lead-poisoned women were likely to abort or delive~ stillborn infants, and articles in the

1880s reported lead to be a teratogen. Because lead passes the

* Indeed, 'lead plasters' were used as abortifacients at the turn of the century. IV-41 placental barrier, the most sensitive population for lead exposure may be fetuses and newborn infants, whose source of exposure to

lead is, of course, the mother. Lead has been implicated in com­ plications of pregnancy, including early and stillbirths, and possibly low-level congenital anomalies. The effects up?n the fetus and neonate are discussed in Section III.C., above. In

this section, we summarize some of the reproductive effects upon women and men, but estimates of populations at risk are made only for women. The Criteria Document (1986; p. 12-192 ff) contains a full discussion of lead's adverse effects upon reproductive function. In addition, the Addendum to the Criteria

Document (1986; p. A-31 ff) contains a section on growth and developmental effects following pre-natal lead exposure, including some studies of negative pregnancy outcomes. Because several early studies (many from the 1800s) showed clear adverse effects of lead at high levels upon female reproduc­

tive functions, particularly miscarriages and stillbirths, women have been largely -- though not entirely -- excluded from occupa­

tional exposure to lead. The mechanisms underlying these effects are unknown at this time. Factors which could contribute range from indirect effects of lead upon maternal nutrition or hormonal

state before or during pregnancy to more direct gametotoxic, embryotoxic, fetotoxic, or teratogenic effects that could affe6t parental fertility or off-spring viability during gestation.

In addition, pregnancy is a stress that may place women at higher risk for lead toxicity, because both iron deficiency

and calcium deficiency increase susceptibility to lead, and IV-42 women have an increased risk of both deficiencies during pregnancy and post-parturition (Rom, 1976). Pregnancy and lactation are also physiological conditions of bone demineralization, when lead as well as calcium and other minerals are released from storage in bones. While this may decrease the total body burden of lead for the pregnant woman, it obviously has potentially toxic consequences for the fetus.

However, there is inadequate information to assess precisely the effects of lead exposure -- at either high or low levels of exposure -- on human ovarian function or other factors affecting female fertility, or on maternal variables, such as hormonal levels, that are known to affect the ability of the pregnant woman to carry the fetus successfully to full term.

While earlier studies focused more upon women, much research is now directed to lead's effect upon male reproductive function.* Lead-related interference with male reproduction function, including gonadal impairment, diminished number and viability of spermatocytes, and apparently exposure-related increases in erectile dysfunction, have been reported. Also, there are several articles implicating exposure of males to lead as the cause of adverse effects on the conceptus (eog., Singhal and Thomas, 1980). These include low fertility rates, low birth weights, and higher rates for miscarriages and stillbirths in families of occupationally lead-exposed men.

* As an indication, the chapter on reproductive effects in Singhal and Thomas (1980) discusses males almost exclusively. IV-43

IV.B.l. Estimating the Population At-Risk for Female Reproductive Effects

While lead is a fetotoxin and therefore probably dangerous to the unborn even at low levels of exposure, the only data on reproductive effects upon the adult female (as opposed to the fetus)* are at fairly high levels, i.e. > 15 ug/dl. No studies have been conducted on reproductive effects of women with 'normal' lead exposures. Recent studies (discussed in the previous chapter) on the inverse relationship between blood lead levels and gesta­ tional age and birth weight and height suggest that reproductive effects of lead exposure observed at high blood-levels continue through the 'normal' range. The available data concerning lead's adverse health effects indicate that the lack of data on reproduc­ tive effects at low exposure levels reflects a lack of data and not a finding of no effect • . To assess the adult female population potentially at risk of suffering reproductive effects, we calculated the number of women of child-bearing age (i.e., aged 15-44) above 15 ug/dl who would benefit from this proposed rule. While there is evidence of neuro­ logical effects, enzymatic inhibition and metabolic alterations at below 10 ug/dl,** this cut-off was used because at 15 ug/dl, many body systems (e.g., heme synthesis) show indications of significant impairment. This estimate should be understood, therefore, as a

* Fetal effects related to lead exposure are discussed in Chapter III.

** The Addendum to the Criteria Document (1986) says, "At present, perinatal blood lead levels at least as low as 10 to 15 ug/dl clearly warrant concern for deleterious effects on early post­ natal as well as prenatal development." (p. A-48). IV-44

low estimate of potential adverse effect. All women of childbearing

age and with blood lead levels over 15 ug/dl were considered to be

at risk of reproductive effects whether or not they were pregnant,

because the damage occurs in any case.

In Section C of Chapter III, estimates are presented of the

number of women of childbearing age, i.e., aged 15 to 44 (24 per­

cent of the total population), and the fraction of them estimated

to have blood lead levels over 15 ug/dl in 1988 (0.36 percent).

Of the total current U.S. population of a little over 240 million,

219 million people are served by community water supplies, of whom 42 million receive water that exceeds a potential MCL of 20

ug/1. Assuming that women of childbearing age and that women with

high blood-lead levels are distributed proportionately throughout

the population,*

219 24% x 240 million x 0.36% x 42 million= 33,000 women in 1988 will be~ 15 ug/dl and, therefore, at risk from

suffering reproductive effects from exposure to lead. By reducing

* While it is reasonable to assume that women of child-bearing age are distributed proportionately throughout the population and therefore that they are equally at-risk of receiving water with high lead levels, it is very conservative to assume that women with high levels of lead in their blood are equally distributed in areas with high water-lead levels and low water-lead levels. This is because blood lead levels are one measure of lead exposurei in general, women with high blood-lead levels are exposed to more lead. Because drinking water is one source of exposure, it is more likely that -- all other sources being equal -- women receiving more lead in their drinking water will have higher than average blood-lead levels. While this is logical, there is no empirical data to calculate the increased likelihood. We have used the most conservative assumption: proportional distribution. IV-45 the contribution of lead from drinking water, this proposed rule will provide benefits to these women in the form of reduced risk of potential reproductive effects.

Also, as presented in Chapter III, to assess the number of pregnant women at risk of suffering adverse effects as a result of exposure to lead from drinking water that exceeds the proposed

MCL, we assumed that pregnant women were distributed proportion­ ately throughout the country and therefore used the national occurrence data to estimate this at-risk population. Of the estimated 54 million women of childbearing age {15-44), about

7 percent are likely to be p~egnant at any given time.* Of these,

680,000 are now probably receiving water that exceeds the proposed

MCL.

IV.C. Monetized Estimates of Adult Health Benefits: Reduced Cardiovascular Disease Risk in Men

Valuing reductions in morbidity and mortality is a diffi­ cult and, to say the least, controversial task. For morbidity, the benefit estimates included avoided medical costs and foregone earnings associated with the diseases. This underestimates social benefits because they fail to account for other important losses associated with disease, including long-term effects, pain and suffering {including, for instance, the paralysis that often follows a stroke). For valuing the reduction in mortality risk, we have chosen a fairly conservative estimate {$1 million per life)

* The rate, according to the Census Bureau, is currently 67.4 pregnancies per 1,000 women of chiid-bearing age. IV-46 from the large range obtained from studies of occupational risk premiums and from other EPA policy papers.

IV.C.1. Hypertension Whether or not it results in coronary or cerebrovascular disease, high blood pressure is a significant chronic illness. It also generates economic costs, in the form of drugs, physi­ cians' visits, hospitalization, and work loss. Data from the NHANES II and from the National Institutes of Health were used to estimate the value of avoiding a case of high blood pressure. The NHANES II ascertained how many times per year a person saw a physician because of high blood pressure. The weighted average, for males 40 to 59 years old with diastolic blood pres­ sure over 90 mm, was 3.27 visits per year. Based upon an average cost of $35 per physician visit, the annual total is $114. The same population was forced to remain in bed an average of 0.41 days per year because of high blood pressure. At the average daily wage ($80),* that translates to $33 per year. Data from the NHANES II also show that 29 percent of those clinically defined as hypertensive were on medication for hyper­ tension. Using standard medical costs indicating that the average drug cost is $220 per year for those on medication yields an annual cost of $64 (in 1985 dollars).

* Based upon wage and earnings data in Statistical Abstracts of the U.S. 1985 and the Economic Report of the President 1986. IV-47

The National Hospital Discharge Survey (1979) found that, excluding those with heart disease or cerebrovascular disease, people with high blood pressure used 3.5 million of the occupied hospital bed-days that year; dividing by the 60 million people· the NHANES II identified as having high blood pressure gives a rate of 0.058 hospital bed-days per ~erson per year. We have assumed that these results apply to the 40 to 59 year old age group of males, as well. Using a daily hospital cost of $450, the annual cost per hypertensive is $26.

Summing these estimates yields a total of $237 per hyper­ tensive per year (1985 dollars). It should be noted that only 29 percent of the people with blood pressure above 90 mm in the

NHANES II were on blood pressure medication, in part because some of them had not previously been detected as having high blood pressure •. Therefore, the average cost for a detected case will be higher. For example, Weinstein and Stason (1977) used an average cost of $200 in 1975 dollars, or about $486 in 1985 dollars, for treatment of patients undergoing medical care for hypertension. Nevertheless, we have conservatively used $250 as the value of avoiding one case of high blood pressure for one year.

IV.C.2. Myocardial Infarctions The estimate of the benefits of reducing the incidence of myocardial infarctions relies heavily on Hartunian et al. (1981), who estimated the medical expenses and lost wages associated with a variety of diseases. Under the category of myocardial infarc­ tions (MI), Hartunian et al. examined three types of cases: IV-48 sudden death, fatal MI, and nonfatal MI. ("Sudden death" was classified as a myocardial infarction in the Pooling Project reQression coefficients.) For each category and each age group, Hartunian et al. obtained data on the type of medical services needed (e.g., ambu­ lance or coronary intensive care unit), the fraction of cases using each service, and the costs in 1975 dollars. They also determined the annualized recurrence and follow-up costs, by age, for each condition. These were then discounted (using a 6 percent real discount rate) to the time of initial occurrence to estimate the cost, in current dollars, of each new case. The resulting estimates were $96 for sudden death and $7,075 for both fatal and nonfatal Mis. These 1975 estimates have been adjusted in three ways to reflect current conditions. First, they are inflated to 1985 dollars. Because most of the costs were hospital-related, with the rest principally being physicians' fees, we inflated the Hartunian et al. cost estimates by a weighted average of 80 percent of the change in the Consumer Price Index (CPI) for hospital rooms and 20 percent of the change in the CPI for physicians' charges.* Approximately 90 percent of the Hartunian et al. MI costs were hospital-related, not physicians' fees, and hospital costs rose faster than physicians' fees, lending a down­ ward bias to the estimates.

* Using data from Tabl~s B-56 and following, Economic Report of the President 1986. IV-49

The second adjustment in the 1975 estimates involves changing cost indices. Because cost indices only account for increased costs of the same procedure, in this case principally the initial hospitalization for a heart attack, they do not reflect the cost of new or different procedures. Since 1975, the fraction of people suffering coronary heart disease who subsequently undergo coronary bypass operations has increased substantially. The number of by­ pass operations tripled in seven years, from 57,000 in 1975 to

170,000 in 1982, while the number of cases of coronary heart disease has remained relatively constant (National Centers for

Health Statistics, Hospital Discharge Survey, and unpublished data). Based on the Hartunian et al. data, 7.1 percent of MI cases in 1975 had subsequent bypass operations. Assuming that they shared proportionately in the tripling of the bypass opera­ tion rate, we estimated that an additional 14 percent of Mis now result in a bypass operation. Hartunian et al. estimated the

.cost of bypass operations at $6,700 in 1975 dollars, or $18,200 in 1985 dollars. Adding 14 percent of this cost to the other direct costs yields an estimate of the total direct costs in

1985 dollars of $21,700 for an MI and $260 for sudden death.

The third adjustm~nt involved discount rates. Hartunian et al. used a 6 percent real discount rate to present value the future year costs, whereas this analysis employs a 10 percent discount rate. Fortunately, Hartunian et al. performed sensi­ tivity calculations for other discount rates, including 10 per­ cent. Making all of these adjustments, the costs per case are

$19,600 for an MI and $233 for sudden death. IV-50

Hartunian et al. also obtained data indicating the proba­ bility distribution of cases among the different categories. Of the total number of cases in these three categories, about 22.5 percent were sudden deaths and the remaining 77.5 percent were fatal or nonfatal Mis. Applying those percentages to the medical-cost estimates derived above yields a weighted average of $15,230 per myocardial infarction.

Hartunian et al. calculated the present value of fore- gone earnings based on reduced labor force participation using data on each type of heart disease, broken down by sex and 10-year age categories. Those results are used here, with several modi­ fications. First, foregone earnings for fatal heart attacks are excluded because the reduction in mortality risks is valued separately (see Section IV.C.4., below). Second, we adjusted for the increase in average non-farm compensation from 1975 to 1985, using information from Data Resources, Incorporated; from the U.S. Census Bureau; and from the Economic Report of the President to Congress. Finally, again a discount rate of 10 per­ cent was used, rather than the 6 percent used by Hartunian et al.

The resulting estimates of foregone earnings are $97,000 for heart attack victims under 45; $51,000 for those between 45 and 54; and $24,000 for those over 55. Based on data from the

Pooling Project and NHANES II, 16.1 percent of nonfatal heart attacks in men between 40 and 59 occur in those under 45, 50.9 percent occur in those between 45 and 54, and 33 ~ercent in those 55 and older. Using those percentages yields a weighted average for lost earnings of $49,500 per attack. Combining that earnings IV-51 estimate with the earlier one for medical costs yields a total benefit per myocardial infarction avoided of about $65,000 (in

1985 dollars).

IV.C.3. Strokes The estimates of the benefits of avoiding strokes also rely on Hartunian et al., with similar adjustments. (Unlike myocardial infarctions, the medical cost estimates for strokes were not adjusted to reflect any changes in medical treatment since 1975.)

Table IV-1 presents the estimates for three types of stroke -­ hemorrhagic, infarctive, and transient ischemic attacks (TIA) -­ by age. The averages are based on the distribution of types of strokes and incidence of strokes by age. The overall average is

$48,000 per stroke avoided, in 1985 dollars.

We have been unable to estimate a value for avoiding the loss in quality of life that occurs in stroke victims. This is a significant omission. For example, of the people in the NHANES II who reported having had a stroke in the past, 45 percent suffered paralysis in the face and 13 percent still had at least partial facial paralysis, 54 percent suffered paralysis in at least one arm and 21 percent remained paralyzed, 59 percent had numbness in arms or legs and 28 percent had remaining numbness,

30 percent had vision impairment and 13 percent remained visually impaired, and 50 percent had speech impairment with 22 percent continuing to suffer from speech impairment. While there are no estimates of people's willingness to pay to avoid the risk of these profound injuries, common sense suggests that it is high. IV-52

TABLE IV-1. Benefits of Reducing: Strokes (1985 dollars 12er case)

Tyee of Stroke Medical Foregone Age Ex12enses Earnings Total

Hemorrhagic

35-44 $13,600 $44,300 $57,900 44-54 14,300 28,100 · 42,400 55-64 18,600 11,900 30,500

Infarctive

35-44 19,000 76,700 95,700 45-54 19,600 46,500 66,100 55-64 25,500 15,100 40,600

Transient ischemic attacks

35-44 3,450 1,200 4,650 45-54 3,450 3,325 6,775 55-64 3,450 8,950 12,400

Weighted average $48,000 IV-53 rv.c.4. Mortality Valuing reductions in mortality is highly controversial.

Over the past decade or so, a substantial literature has developed on the subject. Economists are in general agreement that the best conceptual approach to use is the willingness-to-pay (WTP) of the individuals involved. The appropriate value is not the amount that an individual would pay to avoid certain death, but rather the total sum that a large group of individuals would pay to reduce small risks that sum to one: for example, the amount that 10,000 people would pay to reduce a risk to each of them of one in ten thousand.

Several studies have estimated WTP based on implicit tradeoffs between risk and dollars revealed in market transactions. Most of these studies (e.g., Thaler and Rosen, 1976; Smith, 1974 and 1976; Viscusi, 1978) have studied labor markets, based on the premise that, all else bei~g equal, workers must receive higher wages to accept a higher risk of being injured or killed on the job. Such studies typically regress wages on risk and a variety of other explanatory variables (e.g., levels of education required, worker experience, whether or not the industry is unionized, location, and non-risk working conditions). In such regressions, risk might be measured as the number of fatalities per 1,000 workers per yea-r. The coefficient for that variable is then interpreted as the amount of extra wages needed to compensate for a 0.001 risk of death. Dividing the coefficient by the unit of risk yields the estimate of WTP to avoid a statistical death. IV-54

For example, if the coefficient is $500, the estimated WTP is

$500,000 (= $500/0.001).

A few studies have estimated WTP in non-occupational settings. Blomquist (1977), for example, estimated the implicit cost-risk tradeoffs that individuals make in deciding whether or not to take the time to put on seat belts. None of these studies yields definitive answers. All suffer from data limitations (e.g., incomplete information on possible confounding variables and on the extent to which individuals perceive the risks they face). Not surprisingly, given these problems, the studies also yield a wide range of estimates. A recent survey of the literature prepared for EPA found a range of $400,000 to $7 million per statistical life saved (Violette and Chestnut, 1983). Based on that survey, EPA's guidelines

(US-EPA, 1984c) do not attempt to set any specific value, but rather recommend that range. To simplify the presentation of the results, this analysis uses a single value from the lower end of that range, $1 million per statistical life saved. Although we do not present any formal sensitivity analyses on this value, the results show that the net benefits are so large that they would remain positive whatever part of that broad range is used; even at $400,000 per statistical life saved, the estimated benefits greatly exceed the costs.

IV.C.5. Summary of Annual Monetized Benefits of Reduced. Cardiovascular Disease Table IV-2 summarizes the annual monetized benefits of reduc­ ing the numbers of cases of hypertension, myocardial infarctions, IV-55

TABLE IV-2. St.mmary of Amual Monetized Blood-Pressure Related Benefits of Lowered MCL For Sample Year 1988

Unit I Cost Amual Total Benefits Sul:r-Population (1985 Avoided (millions Category Considered dollars) Cases 1985 dollars)

Cases of hypertension males, $250 130,000 $32.5 aged 40-!59 Myocardial infarctions white males, $65,000 240 $15.6 aged 40-59 Strokes white males, $48,000 80 $3.8 aged 40-59 Deaths white males, $1 million 240 $240.0 aged 40-59

TOI'AL $291.9 (millions 1985 dollars) IV-56 strokes, and deaths due to high blood pressure that would result from the proposed lowering of the allowable amount of lead in drinking water. These are limited for several reasons:

(1) The hypertension estimate covers only males aged 40 to 59.

(2) The other estimates cover only white males aged 40 to 59. (3) No value is assigned to reduced pain and suffering associated with hypertension, myocardial infarctions, and strokes. (4) These are the only adult health effects that are monetized. Most importantly, these estimates assume a causal link between blood lead levels and blood pressure and assume that reducing body lead burden can reduce blood pressure. In addition, of course, some readers may quarrel with the value assigned to reduced risk of mortality; we have chosen a single value for convenience, not because any particular value can be defended strongly. Despite these limitations, the estimated annual benefits of this potential rule are large, tot~lling $291.9 million for sample year 1988.

IV.D. Valuing Health Effects: Caveats and Limitations

To begin valuing the health effects that would be avoided as a result of the proposed MCL for lead in drinking water, we have estimated -- for adult males -- the medical costs, lost earnings, and value of lowered mortality risk associated with reducing the number of hypertensives, strokes, and heart attacks (only white males, aged 40-59, were included in the latter two categories). We also estimated the reduction in the number of deaths from all causes (again, only for white males, aged 40-59) resulting from·the lowered MCL. IV-57

The cost-of-illness estimates themselves are low, primarily because, to reduce potential controversy, the calculations rely on many conservative assumptions. For insta~ce, Hartunian's · estimates (used for adult health costs) are based largely upon actual medical practice and not preferred treatment. As a specific example, their panel of medical consultants indicated that only 5 percent of stroke victims would receive anti-coagulant drugs, less than 5 percent would receive any vocational rehabili­ tation, and that most would receive little or no physical therapy. The real (social) cost of the illness does not decrease if not all victims receive the treatment they need; assuming the treat­ ments are efficacious, stroke victims who are left disabled incur a cost at least equal to the cost of the .medication they should have (but did not) receive. The health benefit estimates, therefore, should be understood as very low lower-bounds for these categories of effects.

Cost-of-illness calculations were not conducted for most of the adverse health effects associated with human exposure to lead including the reproductive effects in both males and females discussed qualitatively in Section C. Among the many other effects not valued monetarily in this health benefit analysis are: - kidney effects, detectable in children at blood lead·

concentrations of about 10 ug/dl, although the damage is

often not manifest until adulthood; - hematopoietic damage, detectable in children at levels

below 10 ug/dl; - adverse pregnancy and other reproductive effects in women,

no threshold indicated; IV-58

- nervous system effects in adults, with behavioral functions

disturbed at very high levels, a dose-dependent slowing of nerve conduction velocity in occupationally-exposed workers, and peripheral nerve dysfunction at 30-50 ug/dl

{central nervous system effects are detectable in children

at 10 ug/dl)1 metabolic changes, detectable in children at about 12 ug/dl1

- enzymatic inhibition, with no threshold indicated in adults

or children, even below 10 ug/dl1

- all effects on fetuses, although lead crosses the placental

barrier and maternal blood-lead values correlate with

several adverse outcomes, including fetotoxicity at high levels and brain damage at lower levels1

- cardiovascular effects on older men and black males of all ages, which may be dose-dependent with no threshold;

- genetoxic and carcinogenic effects of lead;

- lead's effects upon the immune system1 and - lead's effects upon other organ systems (e.g., gastro- intestinal).

Finally, three serious phenomena of lead's adverse effect upon human health were ignored. First, hematopoietic, metabolic, and enzymatic damage have cascading effects throughout the body, which have not been adequately addressed. Second, many of the specific effects have long-lasting sequelae which are not included.

And last, there is a significantly greater chance of serious

effects later in life, including renal failure and cerebral palsy, even in individuals whose highest detected blood-lead IV-59 level was below that associated with the most severe effects and who did not at the time show evidence of lead toxicity~ this increased risk also was not included.

In addition to all the categories of adverse health effects for which we have not yet been able to quantify benefits at all, the costs of the illnesses that were calculated greatly under­ estimate the real (social) benefit~ of preventing those effects, even for the health categories evaluated. The underestimates occur because some categories of direct costs associated with those effects were excluded, as were all indirect but related costs.

In general, society's willingness-to-pay to avoid a given adverse effect is many times greater than the cost of the illness itself, so cost-of-illness analyses inherently underestimate the benefits of avoiding the adverse effect.* Willingness-to-pay studies indicate that society is usually willing to pay two to ten times the cost of medical treatment, and that in specific circumstances society is willing to pay a hundred or a thousand times the cost of the illness itself in order to prevent its occurrence.

More specifically, in the cost-of-illness analyses, only expenses that are directly related to an individual's medical treatment for the specific symptom being evaluated, at the time the symptom occurs, were included. So, for instance, no costs were ascribed for the possibility of adverse effects from the

* For instance, in general people would be willing to pay more than the price of two aspirins to avoid having a headache. IV-60 medical treatment itself or for the possibility that the specific effect of lead may precipitate or aggravate other health effects

(e.g., heart attack and stroke victims are at increased risk of respiratory illness). Related expenses, such as the costs incurred by partially paralyzed stroke victims in purchasing specially designed appliances or retrofitting their existing possessions, or the costs of adapting the home environments for victims of

CVD were also excluded. Finally, no value was ascribed to the pain and suffering of those affected; this is an especially significant omission because, as an example, about half of stroke victims are permanently incapacitated or paralyzed.

All the indirect but related costs of lead's adverse effect upon human health were also left out. These include work time lost by friends and relatives of the victims (including spouses); medical research related to the prevention, detection, or treatment of the effects of exposure to lead; the development of new pro­ cedures to correct the damage resulting from lead exposure;. decreased future e~rnings for those suffering cognitive damage or physical incapacitation (including behavioral disorders) from lead's adverse effects upon virtually every human system; and the like.

IV.E. Summary of Annual Monetized and Non-Monetized Adult Health Benefits of Reducing Exposure to Lead in Drinking Water This chapter discussed two major physiological effects resulting from exposure to lead: cardiovascular changes in males aged 40 to 59 and reproductive impairment in women of childbearing age. Of these, only the male cardiovascular effects were monetized. IV-61

Estimates of the number of women at-risk of reproductive effects,

as well as the number of fetuses potentially at-risk, were presented

but no monetary value was ascribed to them. Table IV-3 summarizes the annual monetized and non--monetized benefits of a potential reduction in the MCL for lead for one sample year, 1988. IV-62

TABLE IV-3. Sl.IDmary of Annual Monetized and Non-:rronetized Health Benefits of Lowered MCL For Sample Year 1988 Unit Cost Annual Benefits Sub-Population (1985 Cases (millions 1985 Effect Considered dollars) Avoided dollars)

M:INETIZED MALE BLOOD-PRESSURE RELATED EFFECT'S Cases of hypertension males, $250 130,000 $32.5 aged 40-59 Myocardial infarctions white males, $65,000 240 $15.6 aged 40-59 Strokes white males, $48,000 80 $3.8 aged 40-59 Deaths white males, $1 million 240 $240.0 aged 40-59 'IOTAL $291.9

NON-M)NETIZED FEMALE REPRODUCTIVE EFFECTS

Adverse reproductive -wa:nen, NA 33,000 NA effects aged 15-44 >15 ug/dl

{Pregnancies at risk) pregnant women, NA 680,000 NA) (of adverse effects) aged 15-44 ) ( - sane as at-risk ) ) ( fetuses ) CHAPTER V

BENEFITS FROM REDUCED MATERIALS DAMAGE

This chapter contains a discussion of the materials benefits that will result from reducing the occurrence of lead in drinking water. Lead seldom occurs naturally in source waters~* primarily it is leached from the pipes and solder by corrosive water.**

Therefore, reducing the occurrence of lead in public water supplies means reducing the corrosivity of that water.+ Reducing the corrosivity of the water produces materials benefits in the form of decreased corrosion damage, in addition to the decrease in lead. This chapter discusses the characteristics of corrosive water and contains estimates of the potential savings that could accrue to water utilities and to consumers by lessening the corrosivity of their water.

As discussed in Chapter II, the major source of lead contami­ nation of drinking wat~r are the materials of the water distribution

* Concentrations of lead in ground water in the United States are typically low. Lead naturally occurring in surface waters or contributed to water by auto emissions, surface run-off, etc. will generally settle in the sediments before reaching the consumer.

** Corrosion is the deterioration of a substance or its proper­ ties due to a reaction with its environment. In this document, the "substance" that deteriorates is the pipe whether made of metal, asbestos-cement, cement, or plastic and the flux and solder joining the pipes, and the "environment" is water, i.e., we are concerned with internal corrosion. (Pipes and other water treatment equipment can also corrode externally.)

+ An alternative, of course, is to replace all plumbing materials containing lead. This would be extremely expensive, costing probably several hundred billion dollars. (Based upon an estimated average replacement cost of $3,000-$5,000 each for most of the 85 million housing units in the country, and $1,500 for each of the 1-10 million housing units estimated to be likely to have a lead service connection.) V-2 and home plumbing systems. Lead occurs primarily as a corrosion by-product. An analysis of the benefits of reducing lead in public water supplies must, therefore, include the benefits of reducing the ~ource of the contamination, that is, the corrosivi'ty (or aggressiveness) of the water.

While pipes made with lead are often considered the source of lead in drinking water, many studies show that lead solder joints actually contribute considerable amounts, as well. In fact, the data show that newly-installed lead soldered pipes conveying corrosive water may leach much more lead than older lead pipes.

In addition, lead may also leach from brass faucets. These issues are also discussed in Chapter II.

The corrosivity of drinking water is important for two main reasons: aggressive water may create or have adverse health effects and the water may cause the plumbing system to deteriorate. Cor­ rosion also affects the aesthetic quality of the water, by stain­ ing fixtures, •discoloring water (most commonly 'red water'), and causing a bad taste.

Corrosive water can be a health problem* because·it leaches contaminants from the supply pipes and distribution system·. and increases the concentrations of metal compounds in the water.

In addition to lead, the metals cadmium, zinc, copper and iron are used in plumbing materials and occur in drinking water as corrosion by-products.

* The potential relationship between corrosive water and cardio­ vascular disease is discussed in Chapter IV. V-3

Sectio"n V .A. ·contains a discussion of what makes water corrosive: the_characteristics of aggressive water, the chemistry of corrosivity, and corrosion indices. In Section V.B., several analyses of the extent of damage to public drinking water systems from internal corrosion are described, and at the end of the section, these analyses are used to quantify the benefits of reducing the corrosivity of U.S. public drinking supplies. The Safe Drinking Water Act requires EPA to set limits, Maxi­ mum Contaminant Levels (MCLs), for drinking water. The National

Primary Drinking Water Regulations require that these limits be· met at the free-flowing outlet of the ultimate user. Because many metals (including lead) occur in drinking water primarily as corrosion by-products, the degree of corrosivity of a system's water is an important consideration in meeting the MCLs at the

tap. However, it is difficult to predict a water's potential corrosivity. In addition,. much of the problem of corrosion is associated with home plumbing.* EPA has not established guidelines

* Utilities are responsible for the integrity of the distribution system and the .quality of the water delivered to customers. But there are many factors adversely influencing end-use water over which they have no direct control. These include the age and condition of the mains and service connections (many older cities have lead pipes), the fact that consumers want soft water (because it is easier to make suds), local building codes often required the use of lead solder to join copper pipes in construc­ tion (the combination of copper and lead results in galvanic corrosion), the condition of residential plumbing, the age and condition of the solder throughout the system (lead is leached quickly from newly-applied solder), and generally poor monitoring of end-use water. Finally, most utilities have little control over the financial resources available to correct identified problems: both their spending and rate structures are usually regulated either by the local government or by a public utility commission. V-4 for corrosivity. However, EPA did require utilities to monitor and test for corrosivity characteristics using the Langelier Saturation

Index (see Section V.A.4. below), and to identify and report on all materials used within the distribution system by February 1983. The purpose of this one-time event was to identify circum­ stances where corrosion contamination was likely to occur, and to encourage appropriate corrective action.

The 1986 Amendments to the Safe Drinking Water Act included a provision banning the use of materials containing lead in public water systems and in residences connected to public water systems. While the ban is effective immediately, States have up to two years to enforce the ban.

V.A. The Characteristics of Aggressive Water Corrosivity is a complex characteristic of water primarily related to pH, alkalinity, dissolved oxygen, total dissolved solids, hardness, velocity, temperature, and other factors.

All water is corrosive to some degree. How aggressive a water

is depends on its physical and chemical characteristics as well as what substance(s) it comes in contact with -- water that is extremely corrosive to some materials may be less corrosive to others. Usually, corrosion is considered a potential problem only for metals, but non-metallic substances (such as asbestos/ cement or cement-lined pipes) can also deteriorate when in contact with water.

Corrosion occurs because of physical and chemical actions between the plumbing materials and the water. The actual V-5 mechanisms of corrosion are usually a complex and interrelated combination of physical, chemical and even biological factors.

V.A.l. Parameters of Water Affecting Corrosivity

The following discussion of the characteristics of water that affect corrosivity summarizes the more detailed presentations in Internal Corrosion of Water Distribution Systems (AWWA-DVGW,

1985), the Corrosion Manual for Internal Corrosion of Water

Distribution Systems (EPA, 1984; p. 11-16), and Larson (1975).

PHYSICAL CHARACTERISTICS: The two main physical characteris­ tics that affect corrosion are flow velocity (which can either increase or decrease the corrosion rate depending on other properties of the water) and temperature (generally, the higher the temperature, the greater the corrosion rate).

CHEMICAL CHARACTERISTICS: Most of what is called corrosion is caused by chemical or electrochemical actions. Many of the chemical factors affecting corrosion rates are related, and a change in one may change others.

J2!!* is a measure of the concentration of hydrogen ion, H+, in the water, which is important because· H+ is one of the major substances that accepts the electrons given up by a metal when it corrodes. In general, at lower pH levels (< 6.8), most metals will corrode more rapidly than at higher pH levels(> 9.0). How­ ever, under certain conditions corrosion can occur at high pH.

* This definition is based upon the discussions in Schock and Gardels, 1983 and US-EPA, 1984. V-6

Corrosion can also occur throughout the range of 5-9 if no protec­ tive film is pr~sent. pH may not have a strictly linear relation­ ship to lead levels in water. The pH level also affects the formation or solubility of protective films on the inside of a pipe.

Alkalinity is a measure of a water's ability to neutralize acids. In potable water, alkalinity is mostly composed of carbonates, which can neutralize acids, and bicarbonates, which can neutralize bases as well as acids. This property is called

"buffering," and can best be understood as resistance to change in pH. Alkalinity affects a water's ability to form a protective coating of lead or calcium carbonate which is especially important in reducing the dissolution of lead. Water with low alkalinity

(i.e., under 60 mg/1 as CaCO3)·or very high alkalinity(> 150 mg/1) is generally corrosive.

Hardness is caused predominantly by the presence of calcium and magnesium ions and is expressed as the equivalent quantity of calcium carbonate (CaCO3) in the water. Hard waters are generally less corrosive than soft waters if sufficient calcium ions and alkalinity are present to form a protective calcium carbonate lining on the pipe walls. (A thin layer of CaCO3 is desirable, as it keeps the water from direct contact with the pipe and reduces the chance of corrosion. "Scaling" occurs when thick layers of

CaCO3 are deposited. Although the pipe is then protected from corrosion, excessive scaling can reduce the carrying capacity of the system, reduce the efficiency of water heaters, clog water meters, etc.) V-7

Dissolved oxygen (DO) is another substance that accepts the electrons given up by the corroding metal and so allows the cor­ rosion reactions to continue. Oxygen also reacts with hydrpgen released by the cathode (see the following discussion on the electrochemistry of corrosion) and with any ferrous iron ions.

Occasionally, oxygen may react with the metal surface to form a protective coating of the metal oxide.

Chlorine lowers the pH of the water, making it potentially more corrosive. In addition, because chlorine is a strong oxidant, it can increase a water's potential corrosivity. A few studies have also shown a difference in corrosion rates depending upon whether the water is chlorinated or chloraminated.

Chlorides and sulfates may cause metal pipes to pit by reacting with the metals and creating soluble metal ions, thus preventing the formation of protective metallic oxide films. Chloride is about three times more ac½ive in this than sulfate. Higher total dissolved solids (TDS) indicate a high ion concentration in the water, increasing conductivity, which in turn increases the water's ability to complete the electrochemical circuit and to conduct a corrosive current. Other factors include the presence of hydrogen sulfide

(generally accelerates corrosion), silicates and phosphates·

(both of which-can form protective films), and natural color and organic matter (which can either inhibit or encourage corrosion, depending upon other characteristics). V-8

BIOLOGICAL CHARACTERISTICS: Both aerobic and anaerobic bacteria can induce corrosion locally, and many organisms form precipitates with iron.

V.A.2. The Electrochemistry of Corrosivity

Generally, metals are most stable in their natural form, i.e., the form in which they occur in native ores and from which they are extracted in processing. The tendency of a metal to return to its natural state (called "activity") is the primary cause of corrosion. Some metals are more active than others and more easily enter into solution as ions or form various compounds.

Zinc, iron and lead are more active than, for example, copper or stainless steel.

The process by which metals corrode in water is electro­ chemical: when a metal enters a solution as an ion or reacts in water with another element to form a compound, electrons will flow from certain areas on the metal's surface to other areas through the metal. An anode is that part of the metal surface that is corroded and from which electric current flows through the metal to the other electrode. The cathode is the metal surface from which current leaves the metal and returns to the anode through the solution. This completes the circuit. All water solutions will conduct a current, a property measured by "conductivity."

The anode and cathode areas may be right next to each other or in different areas of the pipe, and they can set up a current in the same metal or between two different but connected metals. V-9

V.A.3. Types of Corrosion

There are many types of corrosion, which can be either uniform or non-uniform. Uniform corrosion results in an equal amount of material being lost over an entire pipe surface; except in extreme cases, the loss is often so minor that the service life of the pipe is not adversely effected. On the other hand, non-uniform corrosion attacks smaller, localized areas of the pipe, causing holes, restricted flow, or structural failures. Non-uniform corrosion is a serious problem.

There are five basic types of corrosion. Galvanic corrosion occurs when two different metals or alloys come in contact with each other or are in the same environment (e.g., water). This usually occurs at plumbing joints and connections. Due to the differences in their activity, the more active metal corrodes.

Galvanic corrosion is common in household plumbing where different types of metals are used, for instance, copper pipes are joined to galvanized iron pipe or copper pipes are joined together by lead/tin solder.

Pitting is a damaging, localized, non-uniform corrosion that forms pits or holes in the pipe surface. It actually takes very little metal loss to cause a hole in a pipe wall, and failure can be rapid. Pitting is frequently caused by ions of a more-active metal plating out on the pipe surface. Tuberculation occurs when pitting corrosion products build up at the anode next to the pit. Erosion corrosion (or abrasion) mechanically removes protective films, such as metal oxides and CaC03, which serve as V-10 protective barriers against corrosive attacks. Generally, it results from high flow velocities, turbulence, changes in flow direction, and/or the abrasive action of suspended materials.

Biological corrosion results from a reaction between the pipe material and the by-products of organisms such as bacteria.

Dealloying or selective leaching is the preferential removal of one or more metals from an alloy in a corrosive medium.

V.A.4. Corrosion Indices Several indices have been developed to estimate the corrosion potential of specific waters, but because they generally measure the tendency of a specific water to form a protective coating of calcium carbonate, none of these has been entirely successful in predicting whether or not a water is actually corrosive (Larson,

19757 Hoyt et al., 1979; AWWA-DVGW, 1985; etc). The three most commonly used indices (the Langelier Saturation Index, the Aggres­ sive Index, and the Ryznar Stability Index) consider calcium, alkalinity, and pH as parameters to determine the corrosive tendency of the water. However, corrosivity is a complicated and interrelated function of these three characteristics and many others, and each parameter may independently affect the corrosive tendencies of the water. Consequently, some water may be very corrosive even though the measured indexes indicate relatively non-corrosive conditions, or vice versa. It is generally agreed that these indexes are applicable only within a limited pH range, are dependent upon the presence of calcium and alkalinity, and are most appropriate for the materials for which the index was V-11 developed; they are weakest with waters of relatively low alka­ linity and calcium.

The Langelier Saturation Index (LSI} or Langelier Index (LI}, developed in 1936, is one of the first and most widely used; it expresses the potential of the water to either dissolve or pre­ cipitate calcium carbonate. The LSI is defined as the difference between the measured pH of the water and the pH at which CaC03 would be at saturation concentration. The saturation value of the water with respect to CaC03 depends on its pH, calcium ion con­ centration, alkalinity, temperature, and total dissolved solids, such as chlorides and sulfates; but the LSI focuses particularly on the effect of pH upon the solubility of CaC03. A positive LSI value indicates over-saturation and a negative value indicates an undersaturation of CaC03; a value of zero indicates equilibrium. In other words, a positive LSI indicates a tendency for the water to deposit a protective CaC03 layer on the pipe, and hence impede corrosion. Negative values indicate a water's tendency co dissolve CaC03 from the pipe's interior and, thus, a tendency to be aggres­ sive to the pipe. The index is directional only, not quantitative.

The Aggressive Index (AI} is a simplified version of the ·

LSI developed specifically for asbestos pipes. It assumes _typical values for total dissolved solids and for temperature. The AI is nearly interchangeable with the LSI for most practical purposes. Another common measure is the Ryznar Stability Index (RSI}, which uses the same parameters as the LSI. Other corrosion indices include the Larson, (McCauley} Driving Force, and the Riddick

Corrosion Index. V-12

V.A.5. Plumbosolvency and Other Factors Determining Lead Levels in Drinking Water Chapter II contains a more detailed discussion of the potential contamination of drinking water by lead, the sources of that contamination, and potential human exposure resulting from it. This section briefly summarizes the major factors responsible for the contamination of drinking water by lead.

The lead used in service pipes or as part of lead/tin solder is designed to be structurally relatively resistant to corrosion.

In addition, the corrosion rate can be decreased by a relatively insoluble coating that forms on the surface of the metal. However, the combination of copper piping with tin/lead solder found in most residences produces galvanic corrosion that can yield lead levels one to two orders-of-magnitude higher than expected from the composition of the water alone. Many studies found that lead solder, especially newly-applied solder, used with copper household pipes was sufficient to raise lead levels above the current MCL, even with relatively non-corrosive waters. With lead solder, the age of the solder is the single most important variable affecting solubility. As an example, Sharrett et al. (1982a), studying Seattle -- a city with few lead pipes -­ found that the age of the house (a proxy measure for the age of the plumbing materials, including solder) was the dominant factor for predicting the concentration of lead in the tap water. In homes that were newer than five years old, with copper pipes, the median lead concentration for standing water was 31 ug/1 versus

4.4 ug/1 in older homes. The median lead level was 74 ug/1 in v-13· homes built within the previous 18months. New solder will leach lead even in relatively non-corrosive water-~ whether naturally exhibiting low corrosivity or·treated and it can continue to leach significant amounts for up to five years. Two other factors will affect the rate of lead leaching from lead-soldered joints: the surface area of the lead/tin solder at the joints and the number of joints per length of pipe. The duration of contact need not be long. Britton and

Richards (1981) have shown that, with corrosive water, lead levels in systems with copper plumbing joined with lead solder could rise above 100 ug/1 within 40 minutes of·contact. Oliphant

(1983) has presented evidence that these conditions can produce lead levels one to two orders-of-magnitude higher than expected from equilibrium solubility calculations. Several characteristics of lead piping, mentioned in decreas­ ing order of significance, also influence lead levels in drinking water. The length of the lead pipe, in both the home and the supply lines, can have a positive association with lead levels as can the position of the lead pipe. The ratio of the surface area of lead exposed to the water volume contained is another important variable. The age of the dwelling and the percentage of.lead piping in both the service mains and within the residence were also relevant factors in determining lead levels. The number of occupants of the dwelling is inversely proportional to lead levels, probably because fewer occupants mean the water will, on average, remain in the pipes longer (Department of the

Environment, 1977; Pocock, 1980). V-14

Lead can also leach from copper pipes themselves (Herrera et al., 1981). Specifications for copper pipes usually limit only copper and phosphorus, and copper used for non-drinking water applications is permitted to contain some lead. Copper pipe manufacturers have indicated that copper tubing for water is made from the recycled copper products which could result in the introduction of lead impurities (Herrera et al., 1981).

Although not common, lead impurities can also occur in galvanized pipes and from stabilizers used in plastic pipes. Lead is also used in the production of brass and bronze.

Brass is a copper-zinc alloy, which can contain up to 12 percent lead, and bronze is a copper-tin alloy, which can contain up to 15 percent lead (U.S. EPA, 1982b). Both are relatively corrosion resistant, although several studies document lead leaching from bronze and brass fixtures. Additional analysis of the leaching of lead from these and other materials is needed. With even mildly aggressive water, any amount of lead anywhere in the distribution system or household will contribute lead to the drinking water. Overall, the degree of corrosivity, the length of time in the pipe, the total amount of lead in the plumb­ ing system and the newness of the plumbing are the chief determin­ ants of lead concentrations.

In general, with relatively corrosive waters, lead levels in

'first draw' water can be several times higher than in 'running' samples. With aggressive waters and new solder, however, first­ draw samples can be an order-of-magnitude or more greater. V-15

V.B. Damage to Public Water Systems from Internal Corrosion

Because lead contamination of drinking water occurs most often as a result of the corrosive action of water upon the materials of the public and private plumbing systems, strategies to mitigate this contamination will have to include corrosion control efforts.

Corrosion in water supply distribution systems has economic impacts as well as potentially creating health hazards. It also can affect the aesthetic quality of the water. The corrosion rate within a specific water system is a function of the character of the water, the materials used in the distribution system, and of flow conditions. But notwithstanding the local differences, corrosion is a universal problem: corrosion occurs with all metals currently used in plumbing equipment and construction, and also with asbestos­ cement and cement pipes. Corrosion impartially destroys the distribution system mains, service lines and private household plumbing; reduces the flow .capacity and increases operating costs throughout the distribution system; and causes water loss through leaks and pipe breaks.

Corrosion control treatment will produce substantial benefits in the form of reduced damage to public and private plumbing systems, as well as reducing exposure to lead which will produce the health benefits described in Chapters III and IV.

V.B.l. Occurrence of Corrosive Water in the United States

Data on the extent of aggressive water in the United States are incomplete. The most commonly held profile of the corrosivity of U.S. drinking water relies on data on 600 public supply systems V-16 collected in 1962 by the U.S. Geological Survey (Durfor and

Becker, 1964a). That study identified the Northeast, Southeast and Northwest parts of the country as having relatively soft and aggressive waters.* Of the 26 states in those regions, 17 had very soft water (under 60 mg/1 as CaCO3): Alabama, Connecticut,

Delaware, Georgia, Maine, Maryland, Massachusetts, Mississippi,

New Hampshire, New York, North Carolina, Oregon, Rhode Island,

South Carolina, Vermont, Virginia, and Washington. In 1980, these states had a combined population of 67.7 million o.eople. Figure V-1 presents the USGS state findings.

Also during the early 1960s, the U.S. Geological Survey conducted a survey of the aggressiveness of public water supplies in the 100 largest cities in the country (Durfor and Becker, 1964b).

The profile of water corrosivity in this study correlated fairly well with the state study: the Northeast, Southeast, and North­ west are most at risk of very soft water.

In 1974 and 1975, the National Center for Health Statistics (NCHS) conducted an extensive health examination survey of 4,200 randomly selected individuals representing 3,834 households in 35 geographic areas across the country. This was called the National Health and Nutrition Examination Survey, augmentation survey, or

HANES I, augmentation survey. For each geographic area, the

Bureau of Census selected 120 individuals to provide a "represen­ tative" sample of the U.S. population. EPA participated in this

* While these studies r;>resent "average" data on water by state, it should be noted that water (parameters and quality) varies significantly within states, as well. 1" ----- ..

1-xj 1-..: G') c:: :;tj ttj <: I t--'

t--' \0 °'N> .c:: iJl

. , coCi) 0 I--' f 0 I-' 'JQ -...J I-'• n ll> Hardneas,ln t--' parts per million r::vl lS3 Ii 0-60 <: co ~ -<: 61-:ml 0 ,o . ~l H'i ·~ 121-180 T~ :E:: ll> HA~AIL\) rt ~ Greater than co ·~·.,.,..~ ..,. 180 Ii .. PUERTO RICO •

MAP F.- Hardness offinished public water suppl~s of the United Stat111 and Puerto Rico, 198!. (Average.,weighted by population serml) · ·

SOORCE: Durfor and Becker,Chanical- Quality of Public Water Supplies of the United States and Puerto Rico, 1962. U.S.Geological SUrvey, 1964 V-18 survey to assess the role of drinking water quality and cardio­ vascular disease. As part of this survey, the NCHS collected a 1-guart "grab" sample of water from the kitchen faucet of each participant, which was sent to EPA for analysis. Several studies have presented the data from this survey; unfortunately, the results differ from study to study, and the entire data set is currently being re~analyzed at the University of Pittsburgh. One study (Greathouse and Osborne, 1980) indicates that about one-third of the U.S. population is exposed to very soft water (i.e., under 60 mg/1 as CaCO3) and that the median U.S. drinking water is about 91 mg/l-CaCO3. Another paper (Greathouse and Craun, 1~78) presents mean concentrations at

119 mg/1. A third study, Millette et al. (1979), presented the following distribution of aggressive water provided by·utilities (using the Aggressive Index as a measurement): 16.5% of utilities had highly aggressive water (i.e., AI~ 10.0), 52% of utilities had moderately aggressive water (i.e., AI= 10.0-11.9), 31.5% of utilities had non-aggressive water (i.e., AI~ 12.0). No results were presented by Millette on the distribution of population served by those utilities. However, if the average water system serves 3,650 people, this distribution suggests that 36 million people are exposed to very aggressive water and another 114 million people are exposed to moderately aggressive water.

Hudson and Gilcreas (1976), basing their analysis upon the

U.S. Geological Survey data, estimated that half of the water V-19 distributed in the U.S. is naturally corrosive, and is either untreated or, for whatever reason(s), the treatments are not achieving chemical stability. While the specific data they evaluate are related to the 100 largest cities (Durfor and Becker, 1964b), they assumed that larger systems generally provi~e better water than smaller systems. Hudson and Gilcreas then extrapolated linearly to the rest of the country.

In 1979, the A/C Pipe Producers Association commissioned the

Midwest Research Institute (MRI) to survey the occurrence, economic implications and health effects associated with aggressive waters in public water supply systems. MRI surveyed more than three­ quarters of the largest (i.e., serving over 50,000 people) public drinking water systems in the country and about 10 percent of the medium-size (serving 10,000-50,000) systems. Their results (using the Langelier Saturation Index) are extremely close to those of Millette et al.:* 16% of utilities surveyed had highly aggressive waters (i.e., LSI< -2.0),- 51~5% of utilities surveyed had moderately aggressive waters (i.e., -2.0 0. 0) • They estimated the population exposed only for the utilities they sampled; they did not extrapolate to the rest of the country, or attempt to draw a national profile from their data. However, over 171.1 million people are served by medium and large systems.

* The MRI categories using the LSI correspond directly to Millette et al.'s, who used the AI. V-20

Also in 1979, Energy and Environmental Analysis (EEA), using internal data from 1970, estimated that about 27 percent of the U.S. population or 55 million people in 1970 were exposed to very soft water. This includes most of the population in the northeastern, southeastern and western states. While EEA did not cite the source of their data, their map of soft water areas is very similar to the 1962 U.S. Geological Survey.

In Patterson's 1981 analysis using 1978 data from Culligan dealerships throughout the country,* 7 states (Alabama, Connecticut,

Mississippi, North Carolina, Oregon, Rhode Island, and South

Carolina) had soft water, i.e., under 60 mg/1 as CaCO3, with a combined population in these states of 22.1 million people (1980

Census). This is a low estimate of soft water occurrence because the data come from a company providing water-softening services and represent peep.le with harder-than-average water; indeed, the average water hardness in the Culligan data is significantly higher than other data.** Notwithstanding this bias in the data, the profile of the country presented by these data support the U.S. Geological Survey map of hard and soft water areas in the country.

V.B.2. Corrosion Damage Corrosion can take place at the treatment plant, throughout the distribution system, and in household plumbing, and it has many effects that cost utilities and consumers money. Corrosion results in pipes breaking, damage to meters and storage facilities,

* This data set is described in Chapter II. The use of company names and the presentation of related data does not constitute endorsement of these services. ** This issue is discussed in Section II.B.l., above. V-21 water loss, excess repair and replacement of equipment, water damage from leaks, increased puf1!ping costs due to the reduced hydraulic efficiency of corroded or partially blocked pipes, ,loss of service pressure, and increased operating costs associated with all of these effects. In addition, consumers suffer damage to private hot water heatersr radiators, and faucets. Corrosion products also retard heat transfer for heating or cooling water, increasing both water use and operating costs.

EPA has determined (1977) that, as an upper limit, as much as half the water leaving a treatment plant may be lost before ever reaching the consumer. More conservatively, the National

Science Foundation has estimated that, nationally, 15 percent of the water distributed is lost. Of course, not all of this loss results from corrosive water; some is due to accidents or other naturally occurring events. MRI (1979) calculated ' that 38 percent of all water loss or 6 percent of total water distribution is lost due to corrosion from aggressive waters.

MRI's survey of public water utilities indicated that corrosion-related repairs in utilities with non-aggressive water were about 62 percent of those in utilities distributing relatively aggressive water. Data from other studies show that reducing the corrosivity of the water could ~educe corrosion damage rates by at least 20 percent (Bennett et al., 1979, cited in Ryder, 1980) or even 30-75 percent (Dangel, 1975; Kennedy Engineers, 1978).

These surveys also show that pip~ and equipment replacement and repair due to scaling, leakage or breakage* is the major economic

* Delaying the first break is .important because while the probability of a break increases with age, once a break has occurred, the probability of another one is many times higher. V-22 effect of corrosion in distribution systems and that within this category, pipe replacement is the primary maintenance item.

Studies that present monetary estimates of corrosion damage have generally focused upon either the cost to the utility or the cost to the private residence owner; few have done both. Those studies that have tried to include all the costs of corrosion damage have, by and large, covered smaller geographic areas -- a city or metropolitan area, typically.

V.B.3. Estimating the Annual Costs of Corrosive Water

Projections of the economic impact of corrosive water evidence a wide range of factors of concern, assumptions and methodologies, producing, of course, a wide range of "costs."

However", including all of the components of the problem and converting costs to comparable-year estimates, the assumptions and methodologies in the different studies produce a much nar­ rower range of cost estimates than seems likely from an initial review of the literqture.* The factors that must be considered in calculating the annual benefits of reducing corrosivity include the percentage of corrosion damage that is avoidable by water treatment, the relative damage to public and private plumbing systems, total annual estimates of corrosion damage, and the occurrence of corrosive water in the U.S. For comparability, we have calculated per capita estimates and converted all costs to

1985 dollars using fixed-weighted price indexes from the 1986 Economic Report of the President.

* This is all the more surprising because each of the "averages" (costs, damages, water characteristics, etc.} is the mean of a distribution of rather large-~range values. V-23

All water is corrosive. to some degree. Estimates of the

incremental damage from particularly corrosive water and, hence, the potential benefits of corrosivity control treatment vary.

Several studies of the highly aggressive water in Seattle, Washington

(both pre-treatment: e.g., Kennedy Engineers, 1973 and 1978:

Dangel, 1975: and post-treatment, e.g., Courchene and Hoyt,

1985) suggest that water treatment could reduce corrosivity damage by 30-75 percent,* or even more (AWWA-DVGW, 1985).

Hudson and Gilcreas (1976) assumed that corrosive water doubles natural deterioration rates, and Kennedy Engineers (1978) used a 50 percent point estimate of avoidable damage from corrosion.

EEA {1979) and MRI (1979) took a conservatively low point-estimate

{38 percent) from the range presented in the earlier Seattle studies: similarly, Kirmeyer and Logsdon (1983) posit that treat-- ment can reduce corrosivity by 40 percent. Even more conserva­

tively, Ryder {1980) projected that savings from corrosion control would be 25 percent of the total, and Bennett et al. {1979,

cited in Ryder, 1980) estimated that 20 percent of water supply corrosion costs were avoidable.

Estimates of the proportion of total corrosion damage

{maintenance and capital expenses) borne by the private sector**

* Kennedy Engineers suggests that although corrosion damage can be reduced by 30-75 percent, costs can only be reduced by 15-50 percent. This distinction is also made in Internal Corrosion of Water Distribution Systems. {AWWA-DVGW, 1985).

** Of course, cost~ incurred by utilities are eventually passed on to consumers. However, by private sector costs we mean those incurred directly by owners of buildings, and not by the utility. V-24

(primarily homeowners, but also building owners) range from "fully half" (EPA, 1977) to 95 percent (e.g., Dangel, 1975; Ryder,

1980}.* Consumer costs are probably higher than distribution costs for several reasons.

o Residential piping is often composed of copper or

galvanized steel piping joined by brass fittings or

lead/tin solder; combinations of dissimilar metals are particularly vulnerable to galvanic corrosion. o Water used in the home is often heated, increasing its

corrosive potential. o The materials used in residential plumbing are often less

resistant to corrosion and less well-protected than the

materials used in distribution systems (AWWA Committee ' Report, 1984). o Piping in residences is typically smaller than service

mains and flow rates are more variable (both higher and lower), exacerbating· many physical characteristics affecting

corrosion rates.

Three studies set out to calculate only the costs borne by the water utility. Bennett et al •. (1979, cited in Ryder, 1980) used 1975 data from the National Bureau of Standards, which esti­ mated that annual corrosion costs to the overall U.S. economy were $70 billion, of which annual water supply corrosion costs for

* In general, the highest estimates of the proportion of damage borne by the private sector are based upon data from Seattle -­ a city with a relatively new and corrosion-resistant distribu~ tion system. Older cities with less well-protected systems can incur a higher proportion of the total damage, and the total costs will be higher. V-25 distribution systems were $700 million ($1,693.9 million in 1985 dollars). That study assumed that 20 percent of corrosion damage is avoidable through water treatment and estimated the annua-1 benefits to utilities of reducing corrosivity at $1.88 per capita in 1985 dollars. Hudson and Gilcreas {1976) estimated annual savings of $375 million ($782.3 million in 1985 dollars) to com­ muni~y and other public water supplies from treatment to reduce corrosivity; their estimate assumed that aggressive water doubles the natural deterioration rate, decreasing distribution capacity by 2 percent annually, instead of one percent. The per capita annual estimate calculated from this study is $4.34 in 1985 dollars, but it does not include all increased operating costs from corrosion damage. The third study of utilities, Kennedy Engineers, 1973 (cited in Anderson and Berry, 1981), estimated the annual per capita damage from Seattle's highly corrosive water at $2.21 ($5.57 in 1985 dollars). Assuming that 30 percent of those damages were avoidable costs, this yields annual benefits of $1.67 per capita for avoided damage to utility systems.

If utility costs are half of private costs,* these three annual per capita estimates of avoidable damage to utilities {in increasing order, $1.67, $1.88 and $4.34, in 1985 dollars) would yield total annual per capita benefits of $5.01, $5.64 and $13.02 (1985 dollars) for reducing corrosivity.

* That is, 1) total costs of corrosion damage= cost to utilities and cost to private sector, and cost to utilities= 1/2 cost to private sector, 2) benefits of corrosion control= avoidable damage (% decrease in damage) x total costs of corrosion. V-26

Two studies (Kennedy Engineers, 1978~ and Kirmeyer and

Logsdon, 1983) calculated the savings that would accrue to private residence owners from treatment for aggressive water.

Kennedy Engineers, studying Seattle, estimated that the cost of corrosivity damage could be reduced by 20 percent, and

calculated that owners would save $7.50 per year per residential unit ($13.57 in 1985 dollars) if Seattle's highly aggressive water was treated. Using demographic data in the article, this yields annual per capita benefits of $6.17. Kirmeyer and

Logsdon, using a 'typical' situation, assumed that corrosivity

control could reduce damage by 40 percent, yielding present

value benefits of $244 per unit ($292.59 in 1985 dollars) over

the remaining life of the plumbing. Using data from the article

and the AWWA, this yields annual per capita benefits of $9.44

(1985 dollars). If private costs are two-thirds of total

costs (i.e., double the costs -- or benefits -- to utilities), these two per capita annual estimates of benefits to residential

owners ($6.17 and $9.44, in 1985 dollars) would yield total

avoidable corrosion damage benefits of $9.26 and $14.16 per capita per year, in 1985 dollars. Finally, two other studies, Energy and Environmental Analysis

(1979) and Ryder (1980), estimated total savings from treating

water to reduce its corrosion potential. EEA, using data from

Dangel (1975) and Kennedy Engineers (1973), considered pipe damage

to both the public and private ·sectors a·nd calculated potential

annual savings of $2.67 per capita ($4.54 in 1985 dollars). This V-27 is an admitted underestimate because they did not include total increased operating costs {to either the public or private sector), damage to residential hot water heaters or utility equipment other than pipes, etc., and they assumed that residential damage rates are equal to utility rates~ their assumed corrosion rate is also lower than other studies. Ryder, using data collected while he was associated with Kennedy Engineers, calculated total annual corro­ sion damage in Seattle at $7.4 million ($11.7 million in 1985 dollars),* of which 25 percent could be avoided by water treatment~ this yields potential benefits of $5.84 per person per year

(1985 dollars) from control of Seattle's highly corrosive water.

For comparison, Mullen and Ritter (1980) published results on efforts by the Middlesex Water Company in Woodbridge, New Jersey, to reduce damage to their unlined cast iron water mains from their soft and aggressive water. Those treatment efforts were rewarded by reductions in corrosion rates of 70-80 percent, averaged over a 5-6 year period. Alternatively, Hahin (1978) assessed corrosion damage as a function of total operating expenses. His analysis of four Air Force and three Army bases showed that corrosion costs averaged 8-25 percent of total annual operating costs over a 10-year period.

* This is also somewhat of an underestimate {and Ryder presented it as such) because it does not include costs to the suburban water agencies who buy and use Seattle's water, the costs of deterioration of copper pipe, or the costs associated with water conditioning or treatment to minimize corrosion in industrial and institutional buildings. Ryder estimates that the total cost probably exceeds $10 million ($15.8 million in 1985 dollars). Because our costs are all per capita, we would then have to divide this larger figure by the population of the entire ~etropolitan area. V-28

We have not used cost estimates from two studies: MRI (1979) and Anderson and Berry (1981). In the MRI study, the self-reported expenditures attributable to corrosion totalled $8.3 million annually for the utilities surveyed. (Note that this is not a national estimate and that it is an estimate of expenditures, not damage.) The largest estimate in that survey was much higher than any of the others, indicating either more serious problems for that utility or that they misinterpreted the question.

Eliminating that utility drops the estimated costs to $3 million annually for the surveyed utilities. The corresponding per capita costs are $1.15 (including all the utilities) or $0.42

(omitting the outlier); converting to 1985 dollars yields $1.96 and $0.71, respectively. However, the question asked in the survey, "What are your annual costs due to corrosion?", could easily lead to an underestimate of real costs, for several reasons.

First, the utilities reported only on their expenditures and not on those incurred by consumers (which could be much greater).

Second, the utilities reported only on their expenditures and expenditures are a poor estimate of damage; the utilities did not quantify the damage that was occurring (for example, leaking but not broken pipes) but for which they were not (yet) paying

(specifically identified) money. Finally, it is unclear from the data whether any of the utilities identified increased operating costs associated with corrosive water ~for example, water loss from leaks or the increased energy costs of pumping water through pipes partially clogged with.corrosion by-products) or the proportion of regular maintenance costs (e.g. leaks and breaks) that are attributable to corrosion damage. From the V-29

information presented, it is likely that only major capital

expenditures were included.

Anderson and Berry (1981) evaluated the costs and benefits of regulating corrosive water~ They used the arithmetic mean

($2.37 per capita) of the EEA and Hudson and Gilcreas studies

cited above, $2.67 and $2.08 per capita respectively, as their

estimate of the materials benefits of reducing the corrosivity

of drinking water. However, none of the monetized estimates

in the Anderson and Berry article were first converted to same­ year dollars, so they are not comparable. Furthermore, the EEA

estimate includes both private and distribution costs while the

Hudson and Gilcreas analysis considers only distribution costs;

again, the estimates are not comparable. Finally, Anderson and Berry didn't estimate an independent measure of benefit, but relied upon previous work. Because we have included the material

they cite, their analysis offered no independent and additional data.

The range of estimated benefits from treatment to reduce

the corrosivity of water is, then, from $4.54 (the admitted under­ estimate in EEA, 1979) to $14.16 (Kirmeyer and Logsdon, 1983), both in 1985 dollars, per person per year. Table V-1 summarizes

all the studies.

V.B.4. Monetized Benefits of Reduced Corrosion Damage To calculate the annual benefits of reducing the damage caused by corrosive water, we mulEiplied the number of exposed

people by the per capita estimate of corrosion damage. All costs are converted to 1985 dollars. V-30

TABLE V-1. F.st:ima.tes of Annual Per capita Corrosion Damage (1985 dollars)

Corrosion F.stimated Annual Corrosion Damage Damage (per capita) Avoidable Annual Per 'lllrcugh capita Benefits Studies Distrib.ltion Residential Total Water of Corrosion Assunptions/Notes Svstems Treatment Control Kennedy Engineers Assumed 30% potential reduction (1973) $5.57 -- ~16.71* 30%* $5.01* in corrosion danage and that dis- tribution costs were one-third of total costs. Hudson & Gilcreas They did not include increased (1976) $8.68* - $26.04* 50% $13.02* operating costs. Per capita estimate assumes 200 million people are served by public water systems. Assumed that distri- bution costs were one-third of total costs. Kennedy Engineers They calculated $6. 17 per capita (1978) -- $30.87* $46.30* 20% $9.26* in savings to residence c.wners. Assumed residential costs were two-thirds of total costs. Bennett et al. Assumed that 200 million people (1979) $9.40 -- $28.20* 20% $5.64* are served by public water systems (cited in Ryder, and that distribution costs were 1980) one-third of total costs. Energy & Environ- This is an admitted underestimate: mental Analysis $3.98 $7.97 $11.95 38% $4.54 it includes only danage to pipes (1979) (not danage to water heaters, increased operating costs, etc. ) Ryder (1980) $1.17 $22.19 $23.36 25% $5.84 Ryder ascribed 95% of corrosion damage to private owners. Kinneyer & I.Dgsdon - $23.60* $35.40* 40% $14.16* Assumed residential costs were (1983) two-thirds of total damage. AVERAGE $8. 21 w/oor EEA $8.82

* 'lhese estimates have been calculated by the authors of this paper. Assumptions are noted above. V-31

Estimates of the population at risk of exposure to corro­ sive water in the U.S. range from 22.1 million people (Patterson, 1981) to 75.4 million people (Greathouse and Osborne, 1980). Of the available data, we have used the U.S. Geological Survey data on the occurrence of soft wat4ar. In 1980, there were 67.7 million people living in areas identified by USGS as having soft and aggressive water.*

Assuming that these people are served proportionately by community and non-community water systems, 219 67.7 million x 240 million= 61.8 million people would benefit from actions to reduce the corrosivity of their water.

From Table V-1, we have used $8.50 per capita as a point estimate of potential annual savings benefits- from water treatment to reduce corrosivity. This is the mid-point of the estimates including the EEA underestimate ($8.21) and excluding it ($8.82). Multiplied by the potentially exposed population (61.8 million) yields annual materials benefits from reduced corrosivity of $525.5 million in 1985 dollars. For comparison, estimates of average corrosion treatment costs range from under $1 per person per year to about $5 per person per year. The lowest estimates are data collected from 18 cities in six states now known (by EPA) to be treating their water

* This may somewhat underestimate the real exposure to soft water because many people in hard water areas install water softeners. V-32

to reduce its corrosivity. System size varies from 6,000 to

550,000. Annual capital and operating costs for these cities

range from $0.26 per person per year to $1.28 per person per year. system size was not related to per capita cost.

Another estimate, arguably an upper-bound estimate, is from

Applegate (1986). This article presented post-treatment

requirements, options and associated costs for reverse osmosis

(RO)* product water. RO waters are usually extremely corrosive, with pH typically of 5.5-6.9 (Applegate, 1986). Hardness and

alkalinity are typically low, also. Applegate calculated average

costs for post-treatment of RO product waters for use in municipal drinking water systems. Assuming 100 gallons of water used per

person per day, his estimates yield annual capital and operating

costs for various processes ranging from $1.28 to $3.03 per person per year.

The highest estimate of annual cost is from EPA's cost estimates (US-EPA, 1984a), for treatment costs for small systems (i.e., serving up to 1000 people). The point estimate, averaging costs for pH adjustment, use of corrosion inhibitors, and stabilizing

corrosive water, is a little over $5 per person per year. The range, however, is quite wide and highly sensitive to system size. In

some very small systems (i.e., serving 25-100 people), costs may

be many times higher. To be conservative, we used $3.80 as the point estimate of annual per capita treatment costs. This is the arithmetic mid-

* Reverse osmosis is a technology used primarily for desalinizing sea or other brackish water. V-33 point of the technologies evaluated in US-EPA (1984a), adjusting for number of systems and population served. Multiplying by the

61.8 million people estimated to be receiving corrosive waters produces an annual cost estimate of $234.8 million annually, yielding a benefit-to-cost ratio of over 2:1 for materials benefits, alone. Because the point estimate for treatment costs is probably overestimated, net benefits are probably under­ estimated.

These benefits will not be affected by the 1986 Amendments to the Safe Drinking Water Act, which prohibit the use of materials containing lead in public water systems. The estimates in this chapter (both costs and benefits) are based upon the extent of corrosive water in the country, not the population exposed to lead in drinking water.

REFERENCES

Agerty HA. (1952) Lead poisoning in children. Med Clin North Am; 36: 1587-97.

Abu-zeid HAH. (1979) The water factor and mortality from ischemic heart disease: A review and possible explanations for inconsistent findings. Archives of Environmental Health; 34: 328-336. Ainsworth RG, Bailey RJ, Commins BT, Packham RF, Wilson AL. (1977) Lead in drinking water .. Water Research Centre Technical Report 43; England.

Akers CJ, Fellows R. (1978) Review of the problem of lead in drink­ ing water. Presented at 82nd annual conference of the Scottish Institute of Environmental Health; Scotland. American Water Works Association (AWWA). (1984) Water Quality Divi­ sion Committee report. Determining internal corrosion potential in water supply systems. JAWWA; p. 85-88.

American Water Works Association (AWWA). Research Foundation and DVGW - ForschungsstellE~. (1985) Internal Corrosion of Water Distribution Systems, cooperative research report. Printed by th~ AWWA Research Foundation.

Anderson G, et al. (1984) Development and application of methods for estimating: effects of industrial emission controls on air quality impact of reactive pollutants. U.S. EPA July. Anderson PA. (1984) Metal piping and joining materials and fittings. Presented at the Plumbing Materials and Drinking Water Quality Seminar. Sponsored by EPA; Cincinnati, Ohio.

Anderson R, Berry D. (19Bl) Regulating corrosive water. Water Resources Research; 17(6): 1571-1577.

Angina EE. (1979) Geochemistry in Relation to Cardiovascular Disease. National Academy of Sciences, Washington, DC.

Angle CR and McIntire MS,. (1982) Children, the barometer of environ­ mental lead. In: Barness L (ed): Advances in Pediatrics, Yearbook Medical Publishers; Inc.u Chicago; p. 3-32. Angle CR, .McIntire MS, Swanson MS, Stohs SJ. (1982) Erythrocyte nucleotides in children - increased blood lead and cytidine triphosphate. Pediatric Research; 16: 331-4. -2-

Annest JL, Mahaffey KR, Cos DH, et al. (1982) Blood lead levels for persons 6 months - 74 years of age: United States 1976-1980. National Center for Health Statistics, Advanced Data no. 79.

Annest JL, Pirkle JL, Makuc D, Neese JW, Bayse DD, Kovar MG. (1983) Chronological trend in blood lead levels between 1976 and 1980. New England Journal of Medicine; 308: 1373-7. Applegate LE. (1986) Posttreatment of reverse osmosis product waters. JAWWA; May: p. 59-65. Aub JC, Fairhall LT, Minot AS, Reznikoff P, Hamilton A. (1926) Lead Poisoning in Baltimore, MD. The Williams and Wilkins Company (Medicine Monographs: V.7.). Aub JC. (1935) The biochemical behavior of lead in the body. Journal of the American Medical Association; 104: 87-90.

The Australian Therapeutic Trial in Mild Hypertension: Report by the management committee. (1980) Lancet; 2: 1261-7.

Bailey RJ, Russell PF. (1981) Predicting drinking water lead levels. Environmental Technology Letters; 2: 57-66.

Baksi SN, Hughes MJ. (1982) Regional alterations of brain catechola­ mines by lead ingestion in adult rats. Arch Toxicol; 50: 11-18.

Barltrop D. (1969) Transfer of lead to the human fetus. In: Barltrop and Burland (eds), Mineral Metabolism in Pediatrics. Davis Co., Philadelphia. Barton JC, Conrad ME, Nuby S, Harrison L. (1978) Effects of iron on the absorption and retention of lead. Journal of Laboratory and Clinical Medicine; 92: 536-47.

Battelle. (1982) Examination of the leaching of lead from 50 tin - 50 lead soldered copper plumbing by simulated drinking waters. Topical Report; April. Batuman v, Landy E, Maesaka JK, Wedeen RP. (1983) Contribution of lead to hypertension with renal impairment. New England Journal of Medicine; 309: 17-21. Beattie AD, Moore MR, Goldberg A, et al. (1975) Role of chronic low­ level lead exposure with aetiology of mental retardation. The Lancet; March: p. 589-592. -3-

Beevers DG, Erskine E, Robertson M, Beattie AD, Campbell BC, Goldberg A, Moore MR, Hawthorne VM. (1976) Blood-lead and hypertension. Lancet; 2(7975): 1-3.

Beevers DG, Cruickshank JK, Yeoman WB, Carter GF, Goldberg A, Moore MR. (1980) Blood-lead and cadmium in human hypertension. J Environ Pathol Toxicol; 4~ 251-60. Bellinger D, Leviton A, Needleman H, Nichols M, Rabinowitz M, Waternaux c. (1984) Early sensory-motor development and prenatal exposure to lead. Neurobehavioral Toxicology and Teratology; 6:387-402. Bellinger D, Leviton A, Needleman H, Rabinowitz M, Waternaux c. (1985) A longitudinal study of the developmental toxicity of low-level lead exposure in the prenatal and early postnatal periods. Paper presented at the Heavy Metals in the Environment Conference; Athens, Greece. Bellinger D, Leviton A, Needleman HL, Rabinowitz M, Waternaux c. (1986a) Low-level lead exposure and infant development in the first y~ar. Neuro-behav Toxicol Teratol; 8: 151-61. Bellinger D, Leviton A, Needleman H, Rabinowitz M, Waternaux c. (1986b) Correlates of low-level lead exposure in urban children at 2 years of age. Pediatrics. Benignus VA, Otto DA, Muller KE, Seiple KJ. (1981) Effects of age and body lead burden on CNS function in young children: II. EEG Spectra. Electroencephalog Clin Neurophysiol; 52: 240-8.

Betts PR, Astley R, Raine DN. (1973) Lead intoxication in children in Birmingham. British Medical Journal; 1(5850): 402-6.

Bhalla AK, Amento EP, Clemens T, Halick MF, Krane SM. (1983) Specific high-affinity receptors for 1,25-dihydroxyvitamin D3 in human peripheral blood mononuclear cells: presence in monocytes and induction in T lymphocytes following activation. J Clin Endocrinol Metab; 57: 1308-10.

Biliick IH, Curran AS, Shier DR. (1979) Analysis of pediatric blood lead levels in New York City for 1970-1976. Environmental Health Perspectives; 31: 183-90. -4-

Billick IH, et al. (1982} Predictions of pediatric blood lead levels from gasoline consumption. U.S. Department of Housing and Urban Development.

Birden HH Jr, Calabrese EJ, Stoddard A. (1985} Lead dissolution from soldered joints. JAWWA November; p. 66-70.

Blomquist G. (1977} Valuation of Life: Implications of automobile seat belt use. Ph.D. Dissertation; University of Chicago.

Bornschein RL. (1986} Effects of prenatal and postnatal lead exposure on fetal maturation and postnatal growth. Presented at the Joint EPA-EEC Conference on Lead and Neurotoxicity; Edinburgh, Scotland; September. Brennan MJW, Cantril! RC. (1979} y-Aminolevulinic acid is a potent agonist for GABA auto receptors. Nature; 280: 514-5.

Britton A, Richards WN. (1981} Factors influencing plumbosolvency. Journal of the Institute of Water Engineers and Scientists; July.

Brookshire DS, et al. (1986} The valuation of aesthetic preferences. J Environ Econ Manag; 3: 325-46.

Brown C. (1980} Equalizing differences in the labor market. Quarterly Journal of Economics; 94.

Bryant EC. (1966} Statistical Analyses (2nd Edition}. McGraw Hill; New York.

Bryce-Smith D, Deshpunde RR, Hughes J, Waldron HA. (1977) Lead and cadmium levels in still births. Lancet; 1: 1159.

Bryce-Smith D. (1986} Environmental chemical influences on behavior and mentation. Chemical Society Review; 15:93-123. Budiansky s. (1981) Lead: the debate goes on, but not over science. Environmental Science and Technology; 15(3}: 243-246.

Bull RJ, Stanaszek PM, O'Neill JJ, Lutkenhoff SD. (1975} Specificity of the effects of lead on brain energy metabolism for substrates donating a cytoplasmic reducing equivalent. Environmental Health Perspectives; 12: 89-95. -5-

Bull RJ. (1977) Effects of trace metals and their derivatives on the control of brain energy metabolism. Lee SD ed., Biochemical Effects of Environmental Pollutants. Ann Arbor Science; Ann Arbor, Mi; p. 425-40.

Bull RJ, Lutkenhoff SD, McCarty GE, Miller RG. (1979) Delays in the postnatal increase of cerebral eytoehrome concentrations in lead-exposed rats. Neuropharmacology; 18: 83-92.

Bull RJ. (1980) Lead and energy metabolism. In: Singha! PL, Thomas JA (eds). Lead Toxieity. Urban and Schwarzenberg, Inc.; Baltimore: p. 119-68.

Bull RJ, McCauley DT, Tayler DH, Crofton KM. (1983) The effects of lead on the developing central nervous system of the rat. Neurotoxicology; 4(1): 1-17.

Bureau of Food and Drug Administration, Complianee Program Report of Findings, Total Diet Studies, 1973-1980 reports.

Byers RK. (1959) Lead poisoning, review of the literature and report on forty-five cases. Pediatrics; 23: 585-603.

Calabrese EJ, Tuthill RW. (1978} Elevated blood pressure levels and community drinking water characteristics. Journal of Environmental Science and Health; Al3 (10): 781-802.

-Calpbrese E et al. (ed}. (1985) Inorganies in Drinking Water and Cardiovascular Disease. Princeton Scientific Press; Princeton, NJ.

Carmichael NG, Winder C, Lewis PD. (1981} Dose response relation­ ships during perinatal lead administration in the rat: a model for the study of lead effects on brain development. Toxicology; 21: 117-128. .

Charney E, Kessler B, Farfel M, et al. (1983) Childhood lead poisoning. A controlled trial of the effect of dust-control measures on blood lead levels. New England Journal of Medicine; 309: 1089-1093.

Chesney RW, Rosen JF, DeLuca HF. (1983} Disorders of calcium metabolism in ehildren. Chiumello, Sperling M (eds), Recent Progress in Pediatric Endocrinology. Raven Press; New York, NY; p. 5-24. -6-

Chin D, Karalekas PC. (1984) Lead product use survey of public water supply distribution systems throughout the U.S. Pro­ ceedings of: Seminar on Plumbing Materials and Drinking Water Quality; Sponsored by US EPA; Cincinnati, Ohio; May (appendix).

Chisolm JJ Jr, Harrison HE. (1956) The exposure of children to lead. Pediatrics; 18: 943-58.

Chisolm JJ Jr. (1968) The use of chelating agent in the ·treatment of acute and chronic lead intoxication in children. J Pediatr (St. Louis); 73: 1-38.

Chisolm JJ Jr, Barltrop D. (1979) Recognition and management of children with increased lead absorption. Arch Dis Child; 54: 249-62.

Clark, ARL. (1977) Placental transfer of lead and its effects on the newborn. Postgrad. Med. J. 53: 674-678.

Cohen GJ, Ahrens WE. (1959) Chronic lead poisoning: a review of seven years' experience at the Children's Hospital, District of Columbia. J Pediatr (St Louis); 54: 271-84.

Comstock GW. (1979) Water hardness and cardiovascular disease. American Journal of Epidemiology; 110: 375-400.

Comstock GW. (1985) Water quality and cardiovascular disease: a review of recent studies in Canada and the us. Paper presented at the International Symposium on Geochemistry and Health; London, England; April.

Conrad M, Barton J. (1978) Factors affecting the absorption and excretion of lead in the rat. J of Gastroenterology; 74: 731-740. Cooper we, Gaffey WR. (1974-1975) Mortality of lead workers. In: Cole JF (ed.) Proceedings of the 1974 Conference on Standards of Occupational Lead Exposure; J Occup Med; 17: 100-7. Cooper we. (1981) Mortality in employees of lead production facili­ ties and lead battery plants, 1971-1975. In: Lynam Dr, Piantanida LG, Cole JF (eds.) (1978) Environmental Lead: Proceedings of the Second International Symposium on Environmental Lead Research; Cincinnati, OH, December. Academic Press; New York; p. 111-143.

Council on Env°ironmental Quality (1980) Environmental Quality - 1980. Government Printing Office, December.

Courchene JE, Hoyt BP. (1985) The benefits of corrosion treatment in Seattle. Presented at the California/Nevada Section of the AWWA Seminar on Water Quality and Treatment for Corrosion Con­ trol; California. -7-

Crank J. (1975) The Mathematics of Diffusion. Clarendon Press; Oxford, England.

Craun GF, McCabe LG. (1975) Problems associated with metals in drinking water. Journal of the American Water works Association; 67(11):593.

Crofton KM, Taylor DH, Bull RJ, Snulka DJ, Lutkenhoff SD. (1980) Developmental delays i.n exploration and locomotor activity in male rats exposed to low level lead. Life Science; 26: 823-31.

Dangel RA. (1975) Study of corrosion products in the Seattle water department distribution system, U.S. EPA report 670/2-75-036; Cincinnati, Ohio.

De Kort, WL, Verschoor MIA, Wibowo AA, van Hemmen JJ. (1986) Occu­ pational exposure to lead and blood pressure. A study of 105 workers. Arn J Ind Med.

De la Burde B, Choate MS Jr. (1972) Does asymptomatic lead exposure in children have latent sequelae. Journal of Pediatrics; 81: 1088-91.

De la Burde B, Choate MS Jr. (1975) Early asymptomatic lead exposure and development at school age. Journal of Pediatrics; 87: 638-42.

Delves HT. (1970·) A micro-sampling method for the rapid determination of lead in blood by atomic absorption spectrophotometry. Analyst; 431-3.

Department of the Environment. (1977} Lead in Drinking Water: A Survey in Great Britain. Pollution Paper 12; HM 50; London.

Dietrich KN, Krafft KM, Shukla R, Bornschein RL, Succop PA. (1986) The neurobehavioral effects of prenatal and early postnatal lead exposure. In Schroeder (ed): Toxic Substances and Mental Retardation, Neurobehavioral Toxicology and Teratology, AAMD Monography Series; Washington, DC. Dingwall-Fordyce I, Lane RE. (1963) A follow-up study of lead workers. Brit J of Ind Med; 20: 313-5. Donald JM, Cutler MG, Moore MR. (1986) Effects of lead in the labora­ tory mouse. I. Influence of pregnancy upon absorption, retention, and tissue distribution of radio-labeled lead. Environmental Research; 41: 420-431.

Donaldson w. (1924) The action of water on service pipes. Journal of the American Water Works Association; 11(5): 649. -8-

Dresner DL, Ibrahim NG, Mascarenhas BR, Levere RD. (1982) Modulation of bone marrow heme and protein synthesis by trace elements. Environ Res; 28: 55-66.

ORI. (1983 and 1984) The Data Resources Inc: Review of the D.S. economy. Various issues; ORI; Lexington, MA.

Du Pont. (1986) Du Pont comments on the February 1986 draft docu­ ment. Wilmington, Delaware. Available for inspection at: US-EPA Central Docket Section; Washington, DC; Docket No. ECAO-CD-81-2 IIA.E.C.3.6. Durand D. (1971) Stable Chaos. General Learning Corporation; Morristown, NJ.

Durfor CN, Becker E. (1962) Chemical quality of public water supplies of the U.S. and Puerto Rico; Hydrologic Investigations Atlas HA-200; (1964a) U.S. Geological Survey.

Durfor CN, Becker E. (1972) Public water supplies of the 100 largest cities in the U.S.; Geological Survey Water-Supply Paper 1812 (1964b) U.S. GPO.

Edelstein S, Fullmer CS, Wasserman RH. (1984) Gastrointestinal absorption of lead in chicks: involvement of the cholecalciferol endocrine system. J Nutr; 114: 492-700.

Energy and Environmental Analysis (EEA). (1979) Health and corrosion impact of soft water. Prepared for the A/C Pipe Producers Associa­ tion.

Ennis JM, Harrison HE. (1950) Treatment of lead encephalopathy with BAL (2,3-dimercaptopropanol). Pediatrics; 5: 853-68.

Erickson MM, Poklis A, Gantner GE, Dickinson AW, Hillman LS. (1983) Tissue mineral levels in victims of sudden infant death syndrome. I. toxic metals - lead and cadmium. Pediatric Research; 17: 784-99.

Ernhart CB, Landa B, Schell MB. (1981) Subclinical levels of lead and developmental deficits - a multi-variate follow-up reassess­ ment. Pediatrics 67: 911-919.

Ernhart CB, Wolf AW, Kennard MJ, Filipovich HF, Sokol RJ, Erhard P. (1985) Intrauterine lead exposure and the status of the neonate. Paper presented at the Heavy Metals in the Environment Conference; Athens, Greece.

Ernhart CB, Wolf AW, Kennard MJ, Erhard P, Filipovich HF, Sokol RJ. (1986). Intrauterine exposure to low levels of lead: the status of the neonate. Archives of Environmental Health.

Evis MJ, Kane KA, Moore MR, Parratt JR. (1985) The effects of chronic low lead treatment and hypertension on the severity of cardiac arrhythmias induced by coronary artery ligation in anesthestized rats. Toxicol Appl Pharmacol 80: 235-242. -9-

Expert Committee on Trac:e Metal Essentiality. {1983) Independent peer review of selected studies by Drs. Irchgessner and Reichlmayr­ Lias concerning the possible nutritional essentiality of lead: official report of findings and recommendations of an interdis­ _ciplinary expert review committee. Available for inspection at: U.S. Environmental Protection Agency, Environmental Criteria and Assessment Office; Research Triangle Park, NC. Expert Committee on Pediatric Neurobehavioral Evaluations. {1983) Independent peer review of selected studies concerning neuro­ behavioral effects of lead exposures in nominally asymptomatic children: official report of findings and recommendations of an interdisciplinary expert review committee. Available for inspec­ tion at: U.S. Environmental Protection Agency, Environmental Cri­ teria and Assessment Officer Research Triangle Park, NC. Fahim MS, Fahim z, Hall DG. {1976) Effects of subtoxic lead levels on pregnant women in the state of Missouri. Res Commun Chemical Pathol Pharmacol; 13{2): 309.

Favalli L, Chiari MC, Piccinini F, Roza A. {1977) Experimental investigations on the contractions induced by lead in arterial smooth muscle. Acta Pharmacol Toxicol; 41: 412-420.

Ferris B Jr. {1978) Health effects of exposure to low levels of regulated air pollutants. Journal of the Air Pollution Control Association; 28{5): 482. Fisher RA. (1970} Statistical Methods for Research workers. Hafner Press; New York.

Forbes FF. {1900} A very brief discussion of lead poisoning caused by water which has been drawn through lead service pipe. Journal of the New England Water Works Association; vol. 15. Forthofer RN. {1983) Investigation of non-response bias in NHANES II. American Journal of Epidemiology. Fouts PJ, Page IH. {1942) The effect of chronic lead poisoning on arterial blood pressure in dogs. American Heart Journal; 24: 329-31. Fowler BA. {1978) General subcellular effects of lead, mercury, cadmium, and arsenic. Environ Health Perspect; 22: 37-41.

Fowler BA, Kimmel CA, Woods JS, McConnell EE, Grant LD. {1980) Chronic low-level lead toxicity in the rat: III. an integrated assessment of long-term toxicity with special reference to the kidney. · Toxicol Appl Pharmacol; . 56: 59-77.

Freeman AM III. {1982) Air and Water Pollution Control: A Benefit­ Cost Assessment. John Wiley and Sons; New York. -10-

Freeman R. (1965) Reversible myocarditis due to chronic lead poisoning in childhood. Arch Dis Child; 40: 389-93. Gartside PS. (1985) NHANES II statistical overview and evaluation. Presented at: Lead environmental health. Duke University; North Carolina. Gerking s, Stanley L, Weirick w. (1983) An economic analysis of air pollution and health: the case of St. Louis. Prepared for U.S. EPA, Office of Policy Analysis; July. Glasson WA, Tuesday cs. (1970) Environ Sci Technol; 4: 37. Goldberg AM, Meredith DA, Miller S, Moore MR, Thompson CG. (1978) Hepatic drug metabolism and heme biosynthesis in lead-poisoned rats. British Journal of Pharmacology; 62: 529-36.

Goldman JM, Vander AJ, Mouw DR, Keiser J, Nicholls MG. (1981) Multiple short-term effects of lead on the renin-angiotension system. J Lab Clin Med 97: 251-263. Govoni s, Lucchi L, Batlaini F, et al. (1980) Effects of chronic lead treatment affects dopaminergic control of prolactin secretion in rat pituitary. Toxicol Letters; 20: 237-241.

Goyer RA, Rhyne BC. (1973) Pathological effects of lead. Internal Rev Exp Pathol; 12: 1-77.

Grant LD, Kimmel CA, West GL, et al. (1980) Chronic low-level lead toxicity in the rat II. Effects on postnatal physical and behavioral development. Toxicol Appl Pharmacol; 56: 42-58. Greathouse DG, Craun GF. (1978) Cardiovascular disease study: occurrence of inorganics in household tap water and relationships to cardiovascular mortality rates. In: DD Hemphill (ed), Trace Substances in Environmental Health - XII.; p. 31-39.

Greathouse DG, Osborne RH. (1980) Preliminary report on nationwide study of drinking water and cardiovascular diseases. J of Envir Path and Toxicol; 3: 65-76. Gregory R, Jackson PJ. ('1983) Reducing lead in drinking water. Water Research Centre; Regional Seminars May-June.

Gregory R, Jackson PJ. (1984) Central water treatment to reduce lead solubility. Paper presented at AWWA Annual Conference; June; Dallas, Texas.

Gross-Selbeck E, Gross-Selbeck M. (1981) Changes in operant behavior of rats exposed to lead at the accepted no-effect level. Clinical Toxicology; 18: 1247-56. -11-

Gunter EW, et al. (1981) Laboratory procedures used by the Clinical Chemistry Division, Centers for Disease Control, for the Seeond National Health and Nutrition Examination Survey. Centers for Disease Control; Atlanta, GA. Hackney JD~ (1976} Effects of atmospheric pollutants on human physiologic function. Final report; U.S. EPA. Hahin c. (1978) Predicting the metallic corrosion costs of operating and maintaining buildings and utility systems. Materials Performance; September; p. 31-34. Harlan WR, Landis JR, Schmouder RL, et al. (1985) Relationship of blood lead and blood pressure in the adolescent and adult u.s. population. Journal of the American Medical Association; January 25. Harrison. Principles of Internal Medicine. Ninth Edition •.

Hartunian NS, Smart CN, Thompson MS. (1981} The Incidence and Economic Costs of Major Health Impairments. Lexington Books; Lexington, MA. Harvey P, Hamlin M, Kumar R. (1983) The Birmingham blood lead study. Presented at: annual conference of the British Psychological Society, Symposium on lead and health. Available for inspection at: U.S. Environmental Proteetion Agency, Environmental Criteria and Assessment Office, Research Triangle Park, N.C. Hasselblad V. (1981) Modeling dose response relations for health effects data. Environmetrics 81: Selected Papers, SIAM. Heintz HT, Hershaft A, Horak GC. (1976) National Damages of Air and Water Pollution. Report submitted to U.S. EPA. Hernberg s, Nikkanen J. (1970) Enzyme inhibition by lead under normal urban conditions. Lancet; 1: 63-4.

Herrera CE, Ferguson JF, Benjamin MM. (1982) Evaluating the potential for contaminating drinking water from the corrosion of tin-antimony solder. JAWWA; July: p. 368-675. Herrera CE, Kirmeyer GJ, Hoyt BP. (1981, 1982, 1983) Seattle distribution system corrosion control study. Volumes 1, 2, 3. Seattle Water Department. Prepared by EPA Municipal Environ­ mental Research Laboratory; ORD; Cincinnati, Ohio. Hewitt D, Neri LC. (1980) Development of the 'water story': Some recent Canadian studies. Journal of Environmental Pathology and Toxicology; 4 (2 & 3): 51-63.

Hilton PJ (1986). Cellular sodium transport in essential hyper-. tension. New England Journal of Medicine; 314: 222-229. -12-

Holtzman D, Shen Hsu J. (1976) Early effects of inorganic lead on immature rat brain mitochondrial respiration. Pediatric Res; 10: 70-5.

Holtzman D, Shen H~u J, Desautel M. (1981) Absence of effects of lead feedings and growth-retardation on mitochondriae and microsomal cytochromes in the developing brain. Toxicol Appl Pharmacol; 58: 48-56. Hoppenbrouwers T, et al. (1981) Seasonal relationships of sudden infant death syndrome and environmental pollutants. American Journal of Epidemiology; 113(6).

Horvath SM, Folinsbee LJ. (1979),Effects of pollutants on cardio­ pulmonary function. Report to U.S. EPA.

Hoyt BP, Kirrneyer GJ, Courchene JE. (1979) Evaluating home plumbing corrosion problems. JAWWA, December; p. 720-725.

Hudson HE Jr, Gilcreas FW. (1976) Health and economic aspects of water hardness and corrosiveness. JAWWA; 68: 201-204.

Hudson PJ. (1966) Fitting segmented curves whose joint points have to be estimated. Journal of the American Statistical Association; 61: 873-80.

Huseman CA, Hassing JM, Sibilia MG. (1986) Endogenous dopaminergic dysfunction: A novel form of human growth hormone deficiency and. short stature. J. Clin. Endocr. Metab. 62: 484-490. Hypertension Detection and Follow-up Program Cooperative Group. (1982) The effect of treatment on mortality in mild hypertension. New England Journal of Medicine; 307: 976-80.

Iannacone A, Carmignani M, Boscolo P. (1981) Cardiovascular reactivity in the rat following chronic exposure to cadmium and lead. Ann 1st Super Sanita; 17: 655-60.

ICF. (1984) A survey of the literature regarding the relationship between measures of IQ and income. Prepared for U.S. EPA, Office of Policy Analysis. Innes WB. (1981) Effect of nitrogen oxide emissions on ozone in metropolitan regions. Environmental Science and Technology; 15: 933.

Internal Corrosion of Water Distribution Systems. (1985) Report of the AWWA Research Foundation and the DVGW - .Forschungstelle. -13-

J.M. Montgomery Consulting Engineers. (1982 and 1983) Internal Corrosion Mitigation Study, l?inal Report and Addendum. Prepared for Portland, Oregon Water Bureau.

Jackson PJ, Sheiham I. (1980) Calculation of lead solubility in water. Water Research Centre, Technical Report TR-152.

Jacobson J. (1986) The costs and benefits of tightening the MCL for lead, a case study of Boston, Mass. Masters Thesis; Kennedy School of Government; Harvard University.

Janney A. (1982) The relationship between gasoline lead emissions and blood poisoning in Americans. Prepared for U.S. EPA, Office of Policy Analysis.

Johnson NE, Tenuta K. (1979) Diets and lead blood levels of children who practice pica. Environ Res; 18: 369-376. Journal of Pediatrics (Editorial). (1977) New approaches to screening for iron deficiency; 90: 678.

Kakalik JS, et al. (1981) The Cost of Special Education. Rand Corporation Report N-1791-ED.

Kammholz LP, Thatcher LG, Blodgett FM, Good TA. (1972) Rapid proto­ porphyrin quantitation for detection of lead poisoning. Pediatrics; 50: 625-31. Karalekas PC, Craun GF, Hammonds AF, Ryan CR, Worth DJ. (1975) Lead and other trace metals in drinking water in the Boston metropolitan area. Proceedings of the 1975 AWWA Annual Conference; Minneapolis, MN. Karalekas PC, Ryan CR, Larson CD, Taylor FB. (1978) Alternative methods for controlling the corrosion of lead pipe. Journal of the New England Water Works Association; June.

Karalekas PC, Ryan CR, Taylor F'B. (1982) Control of lead pipe corrosion in the Boston metropolitan area. Proceedings of the 1982 AWWA Annual Conference; Miami Beach, Florida.

Karalekas PC, Ryan CR, Taylor FB. (1983) Control of lead, copper, and iron pipe corrosion in Boston. JAWWA; February. Karalekas PC. (1984) Impact of lead piping and fittings on drinking water quality. Proceedings of a Seminar on Plumbing Materials and Drinking Water Quality. Sponsored by US EPA; Cincinnati, Ohio; May.

Keller CA and Doherty RA. (1980a) Bone lead mobilization in lac­ tating mice and lead transfer to sucking offspring. Toxicol. Appl. Pharmacol. 55: 220-228. Keller CA and Doherty RA. (1980b) Lead and calcium distributions in ·blood, plasma, and milk of the lactating mouse. J.Lab. Clin. Med. 95: 81-89. -14-

Kennedy Engineers. (1973) Seattle Corrosion Study (initial); Tacoma, Washington. Kennedy Engineers. (1978) Internal corrosion study, summary report. Conducted for the city of Seattle, Washington; February.

Khera AK, Wibberley DG, Edwards KW, Waldron HA. (1980) Cadmium and lead levels in blood and urine in a series of cardiovascular and norrnotensive patients. Intern J Environ Studies; 14: 309-312.

Kirkby H, Gyntelberg F. (1985) Blood pressure and other cardio­ vascular risk factors of long-term exposure to lead. Scand J Work Envir Health; 11: 15-19.

Kirrneyer GJ, Logsdon GS. (1983) Principles of internal corrosion and corrosion monitoring. JAWWA; February; p. 78-83.

Klein JP, et al. (1980) Hemoglobin affinity for oxygen during short-term exhaustive exercise. Journal of Applied Physiology; 48.

Kleinbaurn DG, Morgenstern H, Keyser LL. (1982) Epiderniologic Research.

Kobayashi J. (1957) On geographical relationship between the chemical nature of river water and death-rate from apoplexy. Berichte des Ohara Instituts fur landwirtschaftliche, Biologie; 11: 12-21.

Kopp SJ, Baker JC, D'Agrosa LS; Hawley PL. (1978) Simultaneous recording of his bundle electrograrn, electrocardiogram, and sys­ tolic tension from intact modified Langendorff rat heart prepara­ tions. I: effects of perfusion time, cadmium, and lead. Toxicol. Appl. Pharrnacol. 46: 475-487.

Kopp SJ, Glonek T, Erlanger M, Perry EF, Perry HM Jr, Barany M. (1980) Cadmium and lead effects on myocardial function and metabolism. J Environ Pathol Toxicol; 4: 205-27. Krornhout D, Couland CL. (1984) Trace metals and CHD risk indicators in 152 elderly men (the Zutphen Study). Eur Heart J; 5 (Abstr suppl 1): 101.

Krornhout D, Wibowo AA, Herber RF, Dalderup LM, Heerdink H, de Lezenne Coulander C, Zielhuis· RL. (1985) Trace metals and coronary heart disease risk indicators in 152 elderly men (the Zutphen study). American Journal of Epidemiology; 122: 378-385.

Kuch A, Wagner I. (1983) A mass transfer model to describe lead concentrations in drinking water. Water Research; 17 (10): 1303-1307.

Larnola A-A, Joselow M, Yamane T. (1975a) Zinc protoporphyrin (ZPP): a simple, sensitive, fluorornetric screening test for lead poison­ ing. Clin Chern (Winston Salem, NC); 21: 93-7. -15-

Lamola A-A, Piomelli s, Poh-Fitzpatrick MB, Yamane T, Harber LC. (1975b) Erythropoietic protoporphyria and lead intoxication: the molecular basis for difference in cutaneous photosensitivity: II. different binding of erythrocyte protoporphyrin to hemo­ globin. J Clin Invest; 56: 1528-5.

Landis JR and Flegal KM. (1986) Randomization model tests for the regression of blood pressure on blood lead using NHANES II data. American Journal of Epidemiology.

Landrigan PJ, Baker EL, Feldman RG, Cox DH, Eden KV, Orenstein WA, Mather JA, Yankel AJ, von Lindern IH. (1976) Increased lead absorption with anemia and slower nerve conduction in children near a lead smelter. Journal of Pediatrics; 89: 904-10.

Larson TE, King RM. (1954) Corrosion by water at low flow velocity. JAWWA; January. Larson TE. (1955) Report on loss in carrrying capacity of water mains; JAWWA.

Larson TE. (1966) Deterioration of water quality in distribution systems. JAWWA; 58 (10): 1307-1316.

Larson TE, McGurk FF. (1973) Complexes affecting the solubility of calcium carbonate in water. Final report. University of Illinois Water Resources Center. Larson TE. (1976) Corrosion by domestic waters. Illinois State Water Survey, Urbana 1975 (also published in W&SE, p. R-176 to R-183). Larson TE, Sollo FW, McGurk FF. (1976) Complexes affecting the solubility of calcium carbonate in water - Phase II. Final Report.· University of Illinois Water Resources Center. Lassovszky P, Hanson H. (1983) Regulatory and operational aspects of controlling corrosion by-products from water supply distribu­ tion systems. AWWA Conference on Water Quality and Treatment: Advances in Laboratory Technology; Pennsylvania. Lassovszky P. (1984) Effect on water quality from lead and nonlead solders in piping. Heating/Piping/Air Conditioning; October; p. 51-58.

Lauwerys R, Buchet JP, Reels H, Hubermont G. (1978) Placental transfer of lead, mercury, cadmium, and carbon monoxide in women. I. Comparison of the frequency distributions of the biological indices in maternal and umbilical cord blood. Environmental Research; 15: 278-89. Lave L, Seskin E. (1977) Air Pollution and Human Health. Johns Hopkins University Press; Baltimore. -16-

Laxen DPH, Harrison RM. (1977) The highway as a source of water pollution. Water Research; 11 :_ 1-11.

Lead Industries Association. (1982) Solders and Soldering: A Primer. L.I.A., Inc; New York.

Leighton J, Shehadi A, Wolcott R. (1983) Aggregate benefits of air pollution control. Prepared for U.S. EPA, Office of Policy Analysis, by Public Interest Economics Foundation; Washington, DC; June.

Levy RI. (1984) The decline in coronary heart disease mortality: Status and perspective on the role of cholesterol. American Journal of Cardiology; August 27; 54 (4).

Lewis BW, Collin RJ, Wilson HS. (1955) Seasonal incidence of lead poisoning in children in st. Louis. Southern Medical Journal; 48: 298-301.

Lilienfeld AM, Lilienfeld DE. (1980) Foundations of Epidemiology. Oxford U Press, New York.

Lincoln RH. (1984) Written statement before the Senate Committee on Environmental and Public Works in Opposition to Senate Bill 2609.

Lin-Fu JS. (1973) Vulnerability of children to lead exposure and toxicity: parts one and two. New England Journal of Medicine; 289: 1229-33, 1289-93.

Lin-Fu JS. (1982) The evolution of childhood lead poisoning as a public health problem. In: Chisholm JJ and O'Hara DM (eds). Lead Absorption in Children. Urban and Schwarzenberg, Inc., Baltimore; p. 1-10. Litman DA, Corriea MA. (1983) L-tryptophan: a common denominator of biochemial and neurological events of acute hepatic porphyrias? Science; 222: 1031-3.

Lorimer G. (1886) Saturnine gout, and its distinguishing marks. British Medical Journal. Lovell J, Isaac R, Singer R. (1978) Control of lead and copper in private water supplies. Carroll County Health Department; Carroll County, Maryland.

Lovering TG (ed). (1976) Lead in the environment. US Geological Survey paper #957; USGPO.

Lyon TDB, Lenihan JMA. (1977) Corrosion in solder-jointed copper tubes resulting in lead conta~ination of drinking water. British Corrosion Journal; 12(1): 41. -17-

Maessen O, Freedman B, Mccurdy R. (1985) Metal mobilization in home well water systems in Nova Scotia. JAWWA; June; p. 73-80.

Mahaffey KR,,Rader J. (1980) Metabolic interactions: lead, calcium and iron. Annals of the NY Academy of Sciences; 355: 285-297.

Mahaffey ~R. (1981) Nutritional factors in lead poisoning. Nutr Rev; 31: 353-362.

Mahaffey KR, Annest JL, Roberts J, Murphy MS. (1982a) National estimation of blood lead levels: United States (1976-1980). New England Journal of Medicine; 307: 573-9.

Mahaffey KR, Rosen JF, Chesney RW, Peeler JT, Smith CM, DeLuca HF. (1982b) Association between age, blood lead concentrations and serum 1,25-dihydroxycholecalciferol levels in children. American Journal of Clinical Nutrition; 35: 1327-31.

Mahaffey KR, Annest JL. (1986) Association of erythrocyte proto­ porphyrin with blood lead level and iron status in the NHANES II, 1976-1980. Environmental Research; 41: 327-338. Mahaffey KR, Gartside PS, Glueck CJ. (1986) Blood lead levels and dietary calcium intake in 1-11 year old children: NHANES II, 1976-1980~ Pediatrics; 78(2): 257-262.

Manton WI. (1977) Sources of lead in blood: identification by stable isotopes. Archives of Environmental Health; 32: 149-59.

Matlack W F. (1980) Statistics for Public Policy and Management. Duxbury Press; Massachusetts.

McBride WG, Black B?, English BJ. (1982) Blood lead levels and behavior of 400 preschool children. Med Journal Aust; 2: 26-9. McCabe LJ, Symons JM, Lee RD, Robeck GG. (1970) Survey of community water supply systems. JAWWA; 62(11): 680-687.

McCauley PT, Bull RJ. (1978) Lead-induced delays in synaptogenesis in the rat cerebral cortex. Fed Proc Fed Am Soc Exp Bio; 37: 740 .. -18-

McGee D, Gordon T. (1976) The results of the Framingham study applied to four other U.S-based epidemiologic studies of coronary heart disease. The Framingham Study; Section 31. DHEW Pub No. (NIH) 76-1083. National Institutes of Health; U.S. Government Printing Office, Washington, D.c. McMichael AJ, Vimpani GV, Robertson EF, Baghurst PA, Clark PD. (1986) The Port Perie Cohort Study: Maternal blood lead and pregnancy outcome. J Epidemiol Commun Health; 40: 18-25.

Mellins RB, Jenkins CD. (1955) Epidemiological and psychological study of lead poisoning in children. Journal of the American Medical Association; 158: 15-20. Mentzer we. (1973) Differentiation of iron deficiency from thalessemia trait. Lancet; 1: 882. ·

Meranger JC, Subramanian KS, Chalifoux C. (1979) A national survey· for cadmium, chromium, copper, lead, zinc, calcium and magne­ sium in Canadian drinking water supplies. Environmental Science and Technology; 13(6): 707-711.

Meredith PA, Moore MR, Campbell BC, Thompson GG, Goldberg A. (1978) Delta-aminolevulinic acid metabolism in normal and lead-exposed humans. Toxicology; 9: 1-9. Midwest Research Institute. (1979) Occurrence and impacts of aggressive water systems. MRI publication. Milar CR, Schroeder SR,. Mushak P, Delcourt J, Grant LD. (1980) Contributions of the caregiving environment to increased lead burden of children. American Journal Mental Deficiency; 84: 339-44. Milar CR, Schroeder SR, Mushak P, Boone L. (1981) Failure to find hyperactivity in preschool children with moderately elevated lead burden. Journal Pediatric Psychology; 6: 85-95. Miller RG, Greathouse D, Bull RJ, Doerger JU. (1985) Absorption of lead from drinking water with varying mineral content. In: Calabrese E et al., Inorganics in Drinking Water and Cardiovascular Disease; Princeton Scientific Press; New Jersey. Millette JR, Hammonds AF, Pansing MF, Hanson EC, Clark PJ. (1980) Aggressive water: assessing the extent of the problem. JAWWA; 72 (5): 262-66. -19-

Minnesota Department of Health. (1985) Preliminary Study of Community Water Systems. Conducted by the Water Supply and Engineering Section, Minnesota Department of Health.

MITRE Corporation. (1979) The environmental lead problem: An assessment of lead in drinking water from a multi-media prospec­ tive. Prepared for u.s. EPA, Office of Drinking Water; May. Moore MR. (1973) Plumbosolvency of waterso Nature; 243: 222-223.

Moore MR, Meredith PA. (1976) The association of delta aminolevulinic acid with the neurological and behavioral effects of lead exposure. In: Hemphill DD ed, Trace Substances in Environmental Health - x. University of Missouri~ Columbia, MO.

Moore MR, Goldberg A, Pocock SJ, Meredith A, Stewart IM, Macanespie H, Lees R, Low A. (1982) Some studies of maternal and infant lead exposure in Glascow. Scottish Medical Journal; 27: 113-122.

Moore MR. Uptake of lead from water. (1985) In: Calabrese E et al. (ed), Inorganics in Drinking Water and Cardiovascular Disease. Princeton Scientific Press; New Jersey. Mooty J, Ferand DF Jr, Harris P. (1975) Relationship of diet to lead poisoning in children. Pediatrics; 55: 636-639. Morbidity and Mortality Weekly Reports, various issues, 1976-1981, U.S. Public Health Service. Morbidity and Mortality Weekly Reports: Reports of the Lead Poisoning Prevention Screening Program 1973-1981. Morgan JM. (1976) Hyperkalemia and acidosis in lead neuropathy. South Med J; 69(7): 881-3. Moreau T, Orssaud G, Juguet B, Busquet G. (1982) Plomb~mie et pression art~rielle: premiers r~sultats d'une enquete transvers­ able de 431 sujets de sexe masculin. [Blood lead levels and arterial pressure: initial results of a cross sectional study of 431 male subjects.] [Letter] Rev Epidemol Sante Fpulique; 30: 3g5-1. -20-

Moriarity M. (1978) Role of calcium in the regulation of adeno­ hypohysial hormone release. Life Sciences; 23: 184-185. Mouw DR, Vander AJ, Cox J, Fleischer N. (1978) Acute effects of lead on renal electrolyte excretion and plasma renin activity. Toxicol Appl Pharmacol; ~6: 435-447.

Mruk SA. (1984) Plastic piping and piping materials. Presented at the Plumbing Materials and Drinking water Quality Seminar; sponsored by EPA; Cincinnati, Ohio; May. Mullen ED, Ritter JA. (1980) Monitoring and controlling corrosion by potable water. JAWWA; May; p. 286-291.

Multiple Risk Factor Intervention Trial: Risk factor changes and mortality results. Multiple Risk Factor Intervention Trial Research Group. (1982) Journal of the American Medical Association; 248: 1465-77.

Murrell NE. (1984) Summary of impact of metallic solders on water quality. Proceedings of Seminar on Plumbing Materials and Drinking Water Quality. Sponsored by US-EPA; Cincinnati, Ohio; May.

Murrell NE. (1985) Impact of metallic solders on water quality. Presented at the Specialty Conference on Environmental Engineering; American Society of Corrosion Engineers; Boston.

Murrell NE, Holzmacher, Mcclendon. (1982) Lead in drinking water due to lead-tin solder joints utilized in interior residential and other plumbing. H2M Corporation; New York.

Nassau County Department of Health. (1985) Report of Investigation of Drinking Water Contamination by Lead/Tin Solder in Nassau County. New York, New York; August. National Academy of Sciences. (1972) Lead: airborne lead in perspec­ tive. National Academy of Sciences; Washington, DC. (Biologic Effects of Atmospheric Pollutants series.)

National Academy of Sciences. (1977-1982) Drinking Water and Health. Safe Drinking Water Committee. (4-volume study) National Academy of Sciences; Washington, D.C.

National Center for Health Statistics. National Hospital Discharge Survey, various issues. -21-

National Center for Health Statistics. (1981) Plan and Opera- tion of the Second National Health and Nutrition Examination Survey (1976-1980). National Center for Health Statistics. (Vital and Health Statistics Series 1, No. 15).

Ne~dleman HL, Gunnoe C, Leviton A, Reed R, Peresie H, Maher C, Barret P. (1979) Deficits in psychological and classroom per­ formance of children with elevated dentine lead levels. New Englind Journal of Medicine; 300: 689-95.

Needleman HL. (1981) The neurobehavioral consequences of low lead exposure in childhood. Neurobehav Toxicol Teratol; 4: 729-732.

Needleman HL. (1984) Comments on chapter 12 and appendix 12C, Air Quality Criteria for Lead (external review draft #1). Available for inspection at: U.S. Environmental Protection Agency, Central Docket Section, Washington, DC; docket no. ECAO-CD-81-211 A.E.C.1.20.

Needleman HL, Bellinger D. (1984) The developmental consequences of childhood exposure to lead: recent studies and methodological issues. In: Lahey BB, Kazdin AE, eds. Advances in Clinical Child Psychology: v. 7; Plenum Press; New York, NY; p. 195-220.

Needleman HL, Landrigan PJ. (1981) The health effects of low level exposure to lead. Ann Rev Public Health; 2: 277-298. Needleman HL, Leviton A, Bellinger D. (1982) Lead-associated iQtellectual deficit [letter]. New England Journal of Medicine: 306: 367. Needleman HL, Geiger SK, Frank R. (1985) Lead and IQ scores: a reanalysis [letter]. Science; 227: 701-704. Needleman HL, Rabinowitz M, Leviton A, Linn s, Schoenbaum S. (1984) The relationship between prenatal exposure to lead and congenital anomalies. Journal of the American Medical Association; 251: 2956-9. Neff CH. (1984) Impact of copper, galvanized pipe, and fittings on water quality. Presented at the Plumbing Materials and Drinking Water Quality Seminar; sponsored by EPA; Cincinnati, Ohio. -22-

Neff CH, Schock MR. (1985) Corrosion product -- corrosion rate -­ water quality investigation in building systems. Prepared for U.S. EPA, Drinking Water Research Division; Cincinnati, Ohio.

Neff CH, Schock MR, Marden JI. (1987) Relationship between water quality and corrosion of plumbing materials in buildings. Prepared for US-EPA Drinking Water Research Division; Cincinnati, Ohio (in press). New approaches to screening for iron deficiency. Editorial. Journal of Pediatrics; 90: 678. Nielsen K. (1975) Contamination of drinking water by cadmium and lead from fittings. Danish Building Res.ear ch Institute.

Nielsen K. (1976) Dissolution of materials from service pipes and house installations and its sanitary aspects. International Water Supply Association. Amsterdam Conference.

Nielsen K. (1984) Corrosion of soldered and brazed joints in tap water. British Corrosion Journal; 19(2): 57-63.

Nordstrom S, Beckman L, Nordenson I. (1978a) Occupational and environmental risks in and around a smelter in northern Sweden: I. variations in birth weight. Hereditas; 88: 43-46. Nordstrom S, Beckman L, Nordenson I. (1978b) Occupational and environmental risks in and around a smelter.in northern Sweden: III. frequencies of spontaneous abortion. Hereditas; 88: 51-54. Nordstrom S, Beckman L, No~denson I. (1979a) Occupational and environmental risks in and around a smelter in northern Sweden: v. spontaneous abortion among female employees and decreased birth weight in their offspring. Hereditas; 90: 291-296. Nordstrom S, Beckman L, Nordenson I. (1979b) Occupational and environmental risks in and around a smelter in northern Sweden: VI. congenital malformations. Hereditas; 90; 297-302.

Nye LJJ. (1929) An investigation of the extraordinary incidence of chronic nephritis in young people in Queensland. Med J Aust; 2: 145-159. -23-

O'Brien JE. (1976) Lead in Boston: its cause and prevention. Journal of the New England Water Works Association; 90(2): 173-180. Odenbro A, Greenberg N, Vroegh K, Bederka J, Kihlstrom JE. (1983) Functional disturbances in lead-exposed children. Ambio; 12: 40-4.

Oliphant RJ. (1982) Lead contamination of potable water arising from soldered joints. IWSA Congress; Zurich.

Oliphant RJ. (1983) Lead contamination of potable water arising from soldered joints. Water Research Centre Report #125E. England.

Otto DA, Benignus VA, Muller KE, Barton CN. (1981) Effects of age and body lead burden on CNS function in young children. I: slow cortical potentials. Electroencephalog Clin Neurophysiol; 52: 229-39.

Otto DA et al. (1982) Effects of low to moderate lead exposure on slow cortical potentials in young children: two-year follow-up study. Neurobehav Toxicol Teratol; 4: 733-7.

Otto D, Robinson G, Baumann S, Schroeder S, Kleinbaum D, Barton C, Mushak P, Boone L. (1984) Five-year follow-up study of children with low-to-moderate lead absorption: electrophysiological evaluation. Presented at the Second International Conference on Prospective Lead Studies; Cincinnati, OH. Available for inspection at: U.S. Environmental Protection Agency, Environmental Criteria Assessment Office, Research Triangle Park, NC. ·

Paglia DE, Valentine WN, Dahlgren JG. (1975) Effects of low level lead exposure on pyrimidine 5' nucleotidase and other erythrocyte enzymes: possible role of pyrimidine 5' nucleotidase in the pathogenesis of lead induced anemia. Journal of Clinical Inves­ tigation1 56: 1164-9.

Patterson JW. (1978) Corrosion inhibitors and coatings. Proceedings of the AWWA annual meeting; June. Patterson JW. (1981) Corrosion in water distribution systems. Prepared for u.s. EPA, Office of Drinking Water; March. Patterson JW, O'Bri~n JE. (1979) Control of lead corrosion. JAWWA; 71 ( 5 ) : 2 6 4-2,71. -24-

Paul c. (1860) Etude sur l'intoxication lente par les preparations de plomb, de son influence sur le produit de la conception [Study of the effect of slow lead intoxication on the product of concep­ tion]. Arch Gen Med: 15: 513-533. Perino J, Ernhart CB. (1974) The relation of subclinical lead level to cognitive and sensorimotor impairment in black preschoolers. Journal of Learning Disabilities: 7: 616-620. Perkin - Elmer. (1977) Analytical Methods for Atomic Absorption Spectrophotometry Using the HGA Graphic Furnace. Norwalk, Connecticut. Perlstein MA, Attala R. {1966} Neurologic sequelae of plumbism in children. Clinic Pediatric (Philadelphia}; 5: 292-8. Perry HM Jr, Erlanger MW. {1978} Presser effects of chronically feeding cadmium and lead together. In Hemphill DD (ed}, Trace Substances in Environmental Health-XII: Proceedings of University of Missouri's 12th annual conference on trace substances in environmental health; June; University of Missouri-Columbia: Columbia, MO: p. 268-75. Perry HM. (1984} Environmental heavy metals and human cardiovascular disease. In Bell and Doege {eds}, Drinking Water and Human Health. American Medical Association. Petrusz P, Weaver CA, Grant LD, et al. {1979} Lead poisoning and reproduction: effects on pituitary and serum gonadotropins in neonatal rats. Environ Res; 19: 382-391. Piccinini F, Favalli L, Chiari MC. {1977} Experimental investigations on the contractions induced by lead in arterial smooth muscle. Toxicology: 8: 43-51. Piomelli S, Davidow B, Guinee VF, Yound P, Gay G. (1973} The FEP (free erythrocyte porphyrins} test: a screening micromethod for lead poisoning. Pediatrics: 51: 254-9. Piomelli S, Seaman C, Zullow D, Curran A, Davidow B. (1977) Metabolic evidence of lead toxicity in "normal" urban children. Clinic Res; 25: 459A. Piomelli S, Seaman C, Zullow D, Curran A, Davidow B. (1982} Threshold for lead damage to heme synthesis in urban children. Proceeding of the National Academy of Sciences in May: 79: 3335-9. -25-

Piomelli s, Rosen JF, Chisolm JJ, Graef JW. (1984) Management of childhood lead poisoning. Pediatrics; 4: 105.

Pirkle JL, Annest JL. (1984) Blood lead levels 1976-1980 - Reply (letter). New England Journal of Medicine; 310(17): 1125-6.

Pirkle J.L, Schwartz J, Landis JR, Harlan WR. (1985) The relationship between blood lead levels and blood pressure and its cardiovas­ cular risk implications. American Journal of Epidemiology; 121(2): 246-58.

Pocock SJ. (1980) Factors influencing household water lead: a British national survey. Archives of Environmental Health; 35(1): 45-51.

Pocock SJ, Shaper AG, Walker M, Wale CJ, Clayton B, Delves T, Lacey RF, Packam RF, Powell P. (1983) Effects of tap water lead, water hardness, alcohol, and cigarettes on blood lead concentrations. J Epidemiol Comm Health; 37: 1-7.

Pocock SJ, Shaper AG, Ashby D, Delves T, Whitehead TP. (1984) Blood lead concentration, blood pressure, and renal function. British Medical Journal; 289: 872-874. Pocock SJ, Shaper AG, Ashby D, Delves T. (1985) Blood lead and blood pressure in middle-aged men. In: Lekkas TD (ed), International Conference on Heavy Metals in the Environment. September; Athens, Greec:.e. Pocock SJ, Ashby D. (1986) Environmental lead and children's intelli­ gence: a review of recent epidemiological studies. Statistician.

The Pooling Project Research Group. (1978) Relationship of blood pressure, serum cholesterol, smoking habit, relative weight and ECG abnormalities _to incidence of major coronary events: Final report of the Pooling Project. J Chron Dis; 31: 201-306.

Port CD. (1977) A comparative study of experimental and spontaneous emphysema. J Toxicol Environ Health; 2: 589-604.

Pounds JG, Wright R, Morrison D, Casciano DA. (1982a) Effect of lead on calcium homeostasis in the isolated rat hepatocyte. Toxicol Appl Pharmacol; 63: 389-401. Pounds JG, Wright R, Kodell RL. (1982b) Cellular metabolism of lead: a kinetic analysis in the isolated rat hepatocyte. Toxicol Appl Pharmacol; 66: 88-101. Pounas·JG, Mittelstaedt RA. (1983) Mobilization-of cellular ca-45 and lead-210: effect of physiologic stimuli. Science; 220: 308-310. -26-

Pounds JG. (1984) Effect of lead intoxication on calcium homeostasis and calcium-mediated cell function: a review. Neurotoxicology; 5: 295-332.

Pounds JG, Rosen JF. (1986) Cellular metabolism of lead: a kinetic analysis in cultured osteoclastic bone cells. Toxi­ cology and Applied Pharmacology; 83: 531-545. Prentice RC, Kopp SJ. (1985) Cardiotoxicity of lead at various per­ fusate calcium concentrations: functional and metabolic responses of the perfused rat heart. Toxicol Appl Pharmacol: 81: 491-501.

Provenzano G. (1980) The social costs of excessive lead-exposure during childhood. In: Needleman (ed}; Low Level Lead Exposure: The Clinical Implications of Current Research. Raven Press, New York.

Provvedini DM, Tsoukas CD, Deftos LJ, Manolagas SC. (1983} 1,25- dihydroxy-vitamin D3 receptors in human leukocytes. Science; 221: 1181-2.

Public Use Data Tape Documentation, Hematology and Biochemistry. (1976-1980}. Second National Health and Nutrition Examination Survey. Catalog number 5411, U.S. Public Health Service, National Center for Health Statistics.

Quarterman J, Morrison J, Humphries W. (1978) The influence of high dietary calcium and phosphate on lead uptake and release. Environ­ mental Research; 17: 60-67. Rabinowitz MB, Wetherill GW, Kopple JD. (1976) Kinetic analysis of lead metabolism in human health. Journal of Clinical Investigation; 58: 260-70.

Rabinowitz MB, Wetherill GW, Kopple JD. (1977) Magnitude of lead intake from respiration by normal man. Journal of Laboratory and Clinical Medicine; 90: 238-248. Rabinowitz MB, Kopple JD, Wetherill GW. (1980) Effect of food intake and fasting on gastrointestinal lead absorption in humans. Am J Clin Nutr; 33: 1784-1788. Rabinowitz MB, Needleman H. (1982) Temporal trends in the lead concentrations of umbilical cord blood. Science; 216: 1429-1431.

Rabinowitz MB, Needleman H. (1983) Petrol lead sales and umbilical cord blood lead levels in Boston, Massachusetts; Lancet.

Rabinowitz MB, Leviton A, Needleman H. (1984) Variability of blood lead concentrations during infancy. Archives of Environmental Health; 39: 74-77.

Ramirez-Cervantes B, Embree JW, Hine CH, Nelson KM, Varner MO, Putnam RD. (1978) Health assessment of employees with different body burdens of lead. Journal Occupational Medicine; 20: 610-7. -27-

Randall A, et al. (1974) Bidding games for valuation of aesthetic environmental improvements. J Environ Econ Manag; 1: 132-49. Rasmussen H, Waisman DM. (1983) Modulation of cell function in the calcium messenger system. Rev Physiol Biochemical Pharmacol; 95: 111-48.

Report of Commissioners, Appointed by Authority of the City Council to Examine the Sources from Which a Supply of Pure Water May Be Obtained for the City of Boston. (1845) JH Eastburn, City Printer, Boston.

Revis NW, Schmoyer RL, Bull R~ (1982) The relationship of minerals commonly found in drinking water to atherosclerosis and hyper­ tension in pigeons; JAWWA; December; p. 656-659. Richards WN, Moore MR. (1982) Plumbosolvency in Scotland - the problem, remedial action taken and health benefits observed. JAWWA; May; p. 901-918. Richards WN, Moore MR. (1984) Lead hazard controlled in Scotish water systems. JAWWA; August1 p. 60-67. Richet G, Albahary C, Morel-Maroqer L, Guillaume P, Galle P. (1966) Les alterations renales dans 23 cas de saturnisme professionnel. [Renal changes in 23 cases of occupational lead poisoning] Bull Memorial Society Medical Hop Paris; 117: 441-66. Robins JM, Cullen MR, Connors BB, et al. (1983) Depressed thyroid indexes .associated with occupational exposure to inorganic lead. Arch Int Med; 143: 220-224. Roels H, Buchet JP, Lauwerys R, Hubermont G, Braux P, Claeys­ Thoreau F, LaFontaine A, Van Overschelde J. (1976) Impact of air pollution by lead on the heme biosynthetic pathway in school age children. Arch Environ Health; 31: 310-6. Rohn RD, Shelton JE, Hill JR. (1982) Somatomedin activity before and after chelation therapy in lead-intoxicated children. Arch Envir Health; 37: 369-373. Rom WN. (1976) Effects of lead on the female and reproduction: a review. Mt. Sinai Journal of Medicine; 43: 542-552. -28-

Rosen JF. (1983) The metabolism of lead in isolated bone cell populations: interactions between lead and calcium. Toxicol Appl Pharmacol; 71: 101-112.

Rosen JF, Chesney RW, Hamstra A, De Luca HF, Mahaffey KR. (19.80a) Reduction in 1,25-dihydroxyvitamin Din children with increased lead absorption. New England Journal of Medicine; 302: 1128-31.

Rosen JF, Chesney RW, Hamstra A, DeLuca HF, Mahaffey KR. (1980b) Reduction in 1,25-dihydroxyvitamin Din children with increased lead absorption. Brown SS, Davis DS ed. Organ Directed Toxicity Chemical Indices and Mechanisms; Pergamon Press, New York, NY.

Rosen JF, Chesney RW. (1983) Circulating calcitriol concentrations in health and disease. Journal Pediatr (St. Louis); 103: 1-7.

Routh DK, Mushak P, Boone L. (1979) A new syndrome of elevated blood lead and microcephaly. J Pediatr Psycho!; 4: 67-76.

Royal Comission on Environmental Pollution. (1983) Ninth report - lead in the environment. Cmnd 8852, Her Majesty's Stationery Office; London, England.

Rummo JH (1974) Intellectual and behavioral effects of lead poison­ ing in children. Dissertation; University of North Carolina. (Available also on University Microfilms.) Rummo JH, Routh DK, Rummo NJ, Brown JF. (1979) Behavioral and neurological effects of symptomatic and asymptomatic lead-exposure in children. Archives of Environmental Health; 34(2): 120-4.

Ryder RR. (1980) The costs of internal corrosion in water systems. JAWWA; May; p. 267-279.

Ryu JE, Ziegler EE, Nelson SE, Fomon SJ. (1983) Dietary intake of. lead and blood lead concentration in early infancy. Am J Dis Child; 137: 886-891. Saenger P, Markowitz ME, Rosen JF. (1984) Depressed excretion of 6 hydroxycortisol in lead-toxic children. J Clin Endocrinol Metab; 58: 363-367. Samuels ER, Meranger JC. (1984) Preliminary studies on the leaching of some metals· from kitchen faucets. Water Research; 18 (1): 75-80.

Sandstead HH, Orth DN, Abe K, et al. (1970) Lead intoxication: effect on pituitary and adrenal function in man. Clin Res; 18: 76. -29-

Sandstead HH, Stant EG, Brill B, et al. (1969) Lead intoxication and the thyroid. Arch Intern Med; 123: 632-635. Sartor F, Van Beneden P, Rondia D. (1981) Water lead concentrations and measurements of health risk in a soft water drinking area. Presented at the Heavy Metals in the Environment conference, Amsterdam; September.

SAS Users Guide. (~982) Statistical Analysis System, Inc. Cary, NC.

Sassa S, Garnick JL, Garnick S, Kappas A, Levere RD. (1973) Studies in lead poisoning. I: Microanalysis of erythrocyte proto­ porphyrin levels by spectrofluorometry in the detection of chronic lead intoxication in the subclinical range. Biochem Medical; 8: 135-48.

Sauer HI. (1980) Geographic patterns in the risk of dying and associated factors, US 196$-1972. DHHS Publication (PHS) 80-1402; Vital and Health Statistics; GPO Washington, DC. Schock MR. (1980 and 1981) Response of lead solubility to dissolved carbonate in drinking water. JAWWA; 72 (12): 695, December and Errata, 73 (3): 36, March.

Schock MR, Gardels MC. (1983) Plumbosolvency reduction by high pH and low carbonate-solubility relationships. JAWWA; 75 (2): 87-91.

Schroeder HA. (1958) Degenerative cardiovascular disease in the Orient. Journal of Chronic Disease; 8: 312-333.

Schroeder SR, Hawk B, Otto DA, Mushak P, Hicks RE. (1984/1985) Separating the effects of lead and social factors on IQ. In: Bornschein and -Rabinowitz (eds). The SecOnd International Con ference on Prospective Studies of Lead. Envir Research; 38: 144~154. Schroeder SR, Hawk B. (1986) Child care giver and environmental factors related to lead exposure and IQ. In: Schroeder (ed), Toxic Substances and Mental Retardation: Neurobehavioral Toxicology and Teratology; (in press).

Schwartz J, Angle C, Pitcher H. (1986) The relationship between childhood blood lead levels and stature. Pediatrics; 77: 281-288. -30-

Schwartz J, Janney A, Pitcher H. (1984b) The relationship between gasoline lead and blood lead. U.S. EPA, Office of Policy Analysis. Schwartz J. (1985) The relationship between blood lead and blood pressure. Presented at the International Conference on Heavy Metals in the Environment. September; Athens, Greece.

Schwartz J and Otto D. (1987) Blood lead levels, hearing thresholds, and neurobehavioral development in NHANES II chtldren and youth. Archives of Environmental Health -- in press.

Seattle Water Metals Survey. (1978) Prepared by Courchene et al. and presented to the Seattle King County Health and Water Depart­ ments and U.S. EPA, Region 10. Shah BV. {1982) SURREGR: Standard Errors of Regression Coefficients from Sample Survey Data. Research Triangle Park, NC, Research Triangle Institute.

Shaper AG, Packham RF, Pocock SJ. (1980) The British Regional Heart Survey: Cardiovascular mortality and water quality. Journal of Environmental Pathology and Toxicology; 4 (2 & 3): 89-112.

Sharrett AR, Carter AP, Orheim RM, Feinleib M. (1982a) Daily intake of lead, cadmium, copper and zinc from drinking water: the Seattle study of trace metal exposure. Environmental Research; 28: 456-475.

Sharrett AR, Orheim RM, Carter AP, Hyde JE, Feinleib M. (1982b) Components of variation in lead, cadmium, copper, and zinc concen­ tration in drinking water: the Seattle study of trace metal exposure. Environmental Research; 28: 476-498. Sharrett AR, Morin MM, Fabsitz RR, Bailey KR. (1984) Water hardness and cardiovascular mortality. In Bell and Doege (eds): Drinking Water and Human Health; American Medical Association.

Sheiham I, Jackson PJ. (1981) The scientific basis for control of lead in drinking water by water treatment. Journal of the Institute of Water Engineers and Scientists; 35(6): 491-515. Shier DR, Hall A. {1977) Analysis of housing and data collected in a lead-based paint survey in Pittsburgh. National Bureau of Standards.

Shlossman M, Brown M, Shapiro E, Dziak R. (1982) Calcitonin effects on isolated bone cells. Calcif Tissue Int; 34: 190-6. -31-

Silbergeld EK, Adler HS. (1978) Subcellular mechanisms of lead neuro­ toxicity. Brain Research; 148: 451-67.

Silbergeld EK, Hruska RE, Bradley D, Lamon JM, Frykholm BC. (1982) Neurotoxic aspects of prophyrinopathies: lead and succinylacetone. Environmental Research; 29: 459-471.

Silbergeld EK, Hruska RE, Miller LP, Eng N. (1980) Effects of lead in vivo and in vitro on GABAergic neurochemistry. J Neurochem; 34:1TI2-8.

Silbergeld E and Schwartz J. (1986) Mobilization of bone lead in women. Presented at: American Society for Pharmacology and Experimental Therapeutics/Society of Toxicology Conference. Baltimore, Md; August.

Silver W, Rodriguez-Torres R. (1968) Electrocardiographic studies in children with lead poisoning. Pediatrics; 41: 1124-7.

Singh NP. (1979) Intake of magnesium and toxicity of lead: An experimental model. Archives of Environmental Health; 34: 168-172.

Singhal PL, Thomas JA eds. (1980) Lead Toxicity. Urban and Schwartzenberg, Inc; Baltimore. Six Kand Goyer R. (1970) Experimental enhancement of lead toxicity by low dietary calcium. J Lab Clin Med; 76(6): 933-942. Six Kand Goyer R. (1972) The influence of iron deficiency on tissue content and toxicity of ingested lead in the rat. J Lab Clin Med; 79(1): 128-136. Smith CM, De Luca HF, Tanaka Y, Mahaffey KR. (1981) Effect of lead ingestion on functions of vitamin D and its metabolites. J Nutr; 111: 1321-1329. Smith M, Delves T, Lansdown R, Clayton B, Graham P. (1983) The effects of lead exposure on urban child~en: the Institute of Child Health/ Southampton study. United Kingdom Department of the Environment. \ \ Smith RS. (1974) The feasibility of an injury\tax approach to occupational safety. Law and Contemporary Problems; 38: 730-44. Smith RS. (1976) The Occupational Safety and Health Act. Washington, DC: American Enterprise Institute for Public Policy Research. Sorrell M, Rosen JF, Roginsky M. (1977) Interactions of lead, calcium, vitamin D, and nutrition in lead-burdened children. Archives of Environmental Health; p. 160-164. -32-

Standard Methods for the Examination of Water and Waste Water. (1971, thirteenth edition). Published jointly by the American Public Health Association, American Water Works Association, and Waber Pollution Control Federat~on; Washington, D.C. Tera, et al. (1985) Identification of gasoline lead in children's blood using isotopic analysis. Arch Env Health; January. Thaler R, Rosen s. (1976) The value of saving a life: evidence from the labor market. Taleckji NE ed. Household Production and Consump­ tion. Columbia University Press, New York. Thomas HF, Elwood PC, Tuthill C, Morton M. (1981) Blood and water lead in a hard water area. The Lancet; May; p. 1047-1048. Thompson NG, Sosnin HA. (1985) Corrosion of 50-50 tin-lead solder in household plumbing. Welding Journal; April; p. 20-24.

Trefry JH, Metz S, Trocine RP, Nelsen TA. (1985) A decline in lead transport by the Mississippi River. Science; 230: 439-441. Treweek GP, Glicker J, Chow B, Sprinker M. (1985) Pilot-plant simulation of corrosion in domestic pipe materials. JAWWA; October; P• 74-82.32 U.S. Department of Health, Education, and Welfare. (1973a) Limita­ tions of activity and mobility due to chronic conditions. Public Health Service, Vital and Health Statistics Series 10, (96). U.S. Department of Health, Education, and Welfare. (1973b) Preva­ lence of selected chronic respiratory conditions: United States - 1970. Public Health Service, Vital and Health Statistics Series 10; (84).

U.S. Environmental Protection Agency. (1977) Survey of Operating and Financial Characteristics of Community Water Systems. (GPO); Washington, D.C.; April. U.S. Environmental Protection Agency. (1982a) 1982 NCLAN Annual Report. Environmental Research Lab; Corvallis, Oregon. U.S. Environmental Protection Agency. (1982b) Corrosion in Potable Water Systems. Prepared by DW DeBerry et al. of SumX Corporation for the Office of Drinking Water; February.

U.S. Environmental Protection Agency. (1982c) Manual for the Certi­ fication of Laboratories Analyzing Drinking Water. (Criteria, Procedures, Quality Assurance). Office of Drinking Water; October. -33-

U.S. Environmental Protection Agency. (1984) Corrosion Manual for Internal Corrosion of Water Distribution S stems. Prepared by Environmental Science and Engineering, Inc. ed: Singley et al.) for the Office of Drinking Water. u.s. Environmental Protection Agency. (1984a) Technologies and Costs.for the Removal of Lead from Potable Water Supplies. Prepared by Ciccone and Associates for the Office of Drinking Wate~; October 23; Final draft.

U.S. Environmental Protection Agency, Office of Policy Analysis. (1984b) Preliminary Regulatory Impact Analysis of Proposed Rules Limiting the Lead Content of Gasoline; July 23. U.S. Environmental Protection Agency. (1984c) Regulatory Impact Analysis Guidelines; Office of Policy Analysis.

U.S. Environmental Protection Agency. (1985a) Occurrence of Lead in Drinking Water, Food and Air. Prepared by JRB Associates for the Office of Drinking water; January. U.S. Environmental Protection Agency. (1985b) The Costs and Benefits of Reducing Lead in Gasoline. Prepared by Schwartz J, Pitcher H, Levin R, Ostro B, Nichols AL; Office of Policy Analysis.

U.S. Environmental Protection Agency. (1986) Air Quality Criteria Document for Lead. Environmental Criteria and Assessment Office (ORD); March. This includes •r'he Addendum to the Criteria Document, which is appended to Volume 1.

U.S. Environmental Protection Agency. (1986a) Methodology for Valuing Health Risks of Ambient Lead Exposure. Prepared by Mathtec, Inc., for the Office of Air Quality Planning and Standards. u.s! Environmental Protection Agency. (1987) Internal Corrosion in Drinking Water Distribution Systems: A Review of EPA's Research Activities. Preliminary draft. water Engineering Research Labora­ tory (ORD); Cincinnati, Ohio; January. U.S. Office of Management and Budget. (1981) Interim Regulatory Impact Analysis Guidance; p. 4.

V.ictery w, Vander AJ, Markel H, Katzman L, Shulak JM, Germain C. (1982a) Lead exposure, begun in utero, decreases renin and angio­ tension II in adult rats. ProcSoc Exp Biol Med 170: 63-67. Victery w, Vander AJ, Shulak JM, Schoeps P, Juluis s. (1982b) Lead, hype1:-tension, and the renin-angiQtens.in s_y_st~m in rats. Journal Lab Clin Med; 99: 354~62. Victery w, Vander AJ, Schoeps P, Germain c. (1983) Plasma renin is increased in young rats exposed to lead in utero and during nursing. Proc Soc Exp Biol Med; 172: 1-7. -34-

Virnpani GV, Wigg NR, Robertson EF, McMichael AJ, Baghurst PA. (1985) The Port Porie Cohort Study: blood lead concen­ trations and childhood developmental assessment. Presented at: Lead Environmental Health: the current issues. Duke University; North Carolina; May.

Violette, Chestnut. (1983) Valuing Reductions in Risks: A Review of the Empirical Estimates. Viscusi WK. (1978) Labor market valuations of life and limb: empirical evidence and policy implications. Public Policy; 26: 359-86.

Wade WA III, et al. (1975) A study of indoor air quality. J Air Pollut Control Assoc; 25: 930-939. Walker R, Oliphant R. (1982a) Corrosion of lead in drinking water. Anti-Corrosion (a British journal); April; p. 8-11. Walker R, Oliphant R. (1982b) Contamination of potable water from lead-based solders. Anti-Corrosion (a British journal); July; p. 13-15.

Webb RC, Winquist RJ, Victery w, Vander AJ. (1981) In vivo and in vitro effects on lead on vascular reactivity in rats. Arn J Physiol; 24: H211-6. Wedeen RP, Maesaka JK, Weiner B, Lipat GA, Lyons MM, Vitale LF, Joselow MM. (1975) Occupational lead neph~opathy. American Journal of Medicine; 59: 630-41.

Weiss ST, Munoz A, Stein A, Sparrow D, Speizer FE. (1986) The relationship of blood lead to blood pressure in a longitudinal study of working men. American Journal of Epidemiology; 123: 800-808. Weinstein MC, Stason WB. (1977) Allocation of resources to manage hypertension. New England Journal of Medicine; 296(13): 732-9. Weston RS. (1920) Lead poisoning by water and its prevention. Journal of the New England water Works Association; volume 34.

White JM, Harvey DR. (1972) Defective synthesis of A and B globin chains in lead poisoning. Nature (London); 236: 71-3.

Wibberly DG, Khera AK, Edwards JH, Rushton DI. (1977) Lead levels in human placentae from normal and malformed births. Journal of Medical Genetics; 14: 339. Williams BJ, Griffith WH III, Albrecht CM, Pirch JH, Hejtrnancik MR Jr. (1977) Effect~ of chronic lead treatment on some cardiovascular responses to norepinephrine in the rat. Toxicology Appl Pharma­ col; 40: 407-13. -35-

Williams BJ, Goldman D, Hejtmancik MR, Ziegler MG. (1978) Nor­ adrenergic effects of lead in neonatal rat. Pharmacology; 20(3): 186.

Williams BJ, Hejtmancik M. (1979) Time and level of perinatal lead-exposure for development of norepinephrine cardiotoxicity. Res Comm CP; 24(2): 367-76.

Williams BJ, Hejtmancik MR, Abreu M. (1983) Cardiac effects of lead. Fed Proc Fed Am Soc Exp Biol; 42: 2989-2~93.

Winneke G, Brockhaus A, Baltissen R. (1977) Neurobehavioral and systemic effects of long-term blood lead elevation in rats. Arch Toxicol; 37: 247-63.

Winneke G. (1979) Impaired intelligence in children from environ­ mental lead (letter). Mun Med Woe; 121(26): 865. Winneke G, Lilienthal H, Werner w. (1982a) Task dependent neuro­ behavioral effects of lead in rats. Arch Toxicol; Supp 5: 84-93.

Winneke G, Hrdina KG, Brockhaus A. (1982b) Neuropsychological studies in children with elevated tooth-lead concentrations. Part 1: pilot study. Int Arch Occupational and Environmental Health; 51: 169-83. Winneke G, Kramer U, Brockhaus A, Ewers u, Kujanek G, Lechner H, Janke w. (1983) Neuropsychological studies in children with elevated tooth lead concentrations. Part II: extended study. Int Arch Occup Environ Health; 51: 231-52. Winneke G, Beginn U, Ewert T, Havestadt C, Kramer u, Krause c, Thron HL, Wagner HM. (1984) Studie zur erfassung subklinischer bleiwirkungen auf das nervensystem bei kindern mit bekannter pranataler exposition in Nordenham •. [Study on the determination of subclinical lead effects on the nervous system of Nordenham children with known pre-natal exposure.] BGA-Berichte. Winneke G, Beginn u, Ewert T, Havestadt c, Kraemer u, Krause C, Thron HL, Wagner HM. (1985a) Comparing the effects of perinatal and later childhood lead exposure on neuropsychological outcome. Environ Res; 38: 155-167. Winneke G; Brockhaus A, Collet w, Kraemer u, Krause C, Thron HL, Wagner HM. (1985b) Predictive value of different markers of lead­ exposure for neuropsychological performance. In: Lekkas TD, ed. International conference: Heavy Metals in the Environment; September; Athens, Greece, v. 1. CEP Consultants, Ltd; Edinburgh, United Kingdoriti ·p:· 44-47. Wolf AW, Ernhart CB, White cs. (1985) Intrauterine lead exposure and early development. In: Lekkas, TD, ed. International conference: Heavy Metals in the Environment; September; Athens, Greece, v. 2. CEP Consultants, Ltd; Edinburgh, United Kingdom; p. 153-155. -36-

Wolman A. (1986) Notes and comments on Hudson and Gilcreas (1976). Journal of the AWWA; 68: 216-217.

Wong, CS, Berrang P. (1976) Contamination of tap water by lead pipe and solder. Bulletin of Environmental Contamination and Toxicology; 15(5): 530.

Wong GL. (1983) Actions of parathyroid hormone and 1,25-dihy­ droxcholecalciferol on citrate decarboxylation in osteoblast-like bone cells differ in calcium requirement and in sensitivity to · trifluoperazine. Calcif Tissue Int; 35: 426-31.

World Health Organization, United Nations Environmental Program. (1977) Lead: Environmental Health Criteria 3; Geneva, Switzerland.

Worth D, Lieberman M, Karalekas P, Craun G. (1981) Lead in drinking water: The contribution of house tap water to blood lead level. Lyna~ et al. (eds), Environmental Lead. Academic Press, p. 199- 225.

Yanke! A, Von Lindern J, Walters. (1977) The Silver Valley lead study: The relationship between childhood blood lead levels and environmental exposure. J Air Pollut Control Assoc; 27: 763-767.

Yip R, Norris TN, Anderson AS. (1981) Iron status of children with elevated blood lead concentrations. Journal Pediatr (St. Louis); 98: 922-5. Yule w, Landsdown R, Millar IB, Urbanowicz MA. (1981) The relation­ ship between blood lead concentrations, intelligence and attainment in a school population: A pilot study. Dev Medical Child Neurology; 23: 567-76. -

Yule W, Lansdown R. (1983) Lead and children's development: recent findings. Presented at international conference: Management and· Control of Heavy Metals in the Environment; Septemberi Heidelberg, West Germany. Yule w, Urbanowicz MA, Lansdown R, Millar IB. (1984) Teacher's ratings of children's behaviour in rel~tion to blood lead levels. British Journal of Developmental Psychology. APPENDIX A: BOSTON CASE STUDY

THE COSTS AND BENEFITS OF TIGHTENING THE MAXIMUM CONTAMINANT LEVEL FOR LEAD IN DRINKING WATER FROM .05 MG/L TO .01 MG/L: A CASE STUDY OF BOSTON, MASSACHUSETTS

A Policy Analysis Exercise

Submitted in Parti•l Fulfillment of the Requirements for the Masters in Public Policy Degree Kennedy School of Government, Harvard University

Jonathan Jacobson April 14, 1986

TABLE OF CONTENTS

EXEC:UT I VE SUMMARY ••• • •• II •••••••••••••••••• • ••••••••• Page 1

INTf;!QDU1::TION • •• • • • • • • • •., • • • • • • • • • • 11 • 11 • • • • • •• • • • • • • • • Page 3

8A1::•:::13F.:OUND ••••••••••••• II •••••••••••••• D ••••••••••••• Page 5

BOSTON: ITS WATER AND CURRENT TREATMENT ..•••.••••.. Page 1 ·-:,....

PROPOSED TREATMENT AND ITS COST ....••..•...•..•••.. Page 15

HE,~L Tl-f BENEFITS . II ...... II • Page 20

CH I LDf:;;'.EN' S BENEFITS •••••••••••••••••••••••••••••••• F" ,=1.ge :22

HEALTH BENEFITS -- ADULT WHITE MALES, AGES 40-59 ••• Page 26

Hypertension••••·••··•··•••·•••••·••••·•• F'age 28

Cardi ,::ivascL1l ar Disease ••••••••••••••••••• Page 28

My,:11: ar di al I ti far,: ti ,:,tis ••••••••••••••• Page 2·~

St 1r •::alc:es. • • • • • • • • • • • • • • • • c • • • • • • • • • • • • Page 30

Deat~1s .•..••. .•...••...... •.... Page 30

Total Benefits for Adult White Males •.••. Page 31

MA TEF.: I ALS BENEFITS ••••••••••••••••••••••••••••••••• ·Page- 33

RESULTS AND DISCUSSION •••.••.••.•••...... •..•...•. Page 36

F.:EC:OMMENDA TI ONS •••••••••••••••••••••••••••••••••••• P,age 40

EXECUTIVE SUMMARY

This paper reports the findings of the cost/benefit analysis of lowering the maximum contaminant level (MCL) for lead in drinking water to .01 mg/L using the City of Boston as a case stL1dy. After discussing the problem of corrosion related lead contamination in general terms and for Boston in particular, my study estimated the costs of additional corrosion cdntrol required to lower lead levels in Boston to the contemplated standard, the health benefits resulting from the lower lead levels, and· the benefits of reduced materials damage produced by the additional treatment of drinking water.

Although my analysis generated results under different assumptions, the net benefit for the most li~ely case is

$7,200,000 (1985$) in 1989 rising to $8,000,000 in 1992. The

Boston share of treatment costs in 1988 is $700,000. Ot1 the benefits side, the analysis yielded the following estimates in

1 '388:

$340,000 for children's health effects,

$1'3<),000 f,::ir av,:,ided my,::i,:ardial it1far,:ti,:;r,s,

$115,000 for avoided strokes,

$6,300,000 for avoided deaths, and

$940,000 for r~duced matErials damage.

In terms of benefit/cost ratios, the results are 11.5:1 in 1988 rising to 12.7:1 in 1992. Pop~lation growth in Boston accounts for the steady rise in health benefits throughout the period. 2

In order to account for uncertainties and statistical biases r•sulting from both the data employed in the analysis and the exclusion ,:::)f important ,:at~g,:,ries ,::,f health effe,:ts (e.g. retial damage, pregnancy complications, and cardiovascular disease in other age groups and in blacks), sensitivity analysis was performed. Even under a pessimistic set of assumptions, additional treatment would still yield positive net benefits.

Using optimistic assumptions, the case for lowering the MCL for lead is overwhelming (net benefits= $11.5 million in 1988).

Th~ results of the analysis suggest that EPA should take the following actions:

1. Lower the maximum contaminant lever for lead from .05 mg/L to .01 mg/L. The agency should consider waivers in exceptional cases.

:2. Pr,::,vide te-,:hni,:al as::sistan,:e atid information to localities to aid them in their efforts to control corrosion thereby reducing leve-ls of lead and other contaminants in drinking water. 3

INTRODUCTION

Since Roman times, people have known about some of the toxic

effects of human exposure to lead. Despite these concerns, lead

has continued to be used for a variety of purposes: as an

additive in paint and gasoline and as a material for water

conveyance pipes. It was not until recently, however~ that

governments have taken action to reduce environmental exposure to

In the United States, the use of lead based paint was

banned. The federal government also regulated the use of lead as

an additive in gasoline beginning in the e,rly 1970s. Among toxic substances, lead was one of the fir~t for which exposure

standards were established.

Increasingly, research has indicated that physiological and neurophysiological damage can n.~~ult from exposure to levels

previously tho~ght to be s~fe. Ln light of this evidence, the

Centers for Disease Control (CDC) lowered the criteria for lead

poisoning in children from 30 µg/dl to 25 µg/dl (when coupled with free erythrocyte protoporphyrin CPEP) of 35 µg/dl). It has motivated the EPA to propose a phase out of the use of leaded gasoline. In response to the growing concern about low-level exposure to lead, EPA also has proposed that the MCL for lead be lowered to .01 mg/L.

While lead can be found in both ground and surface water supplies, the major source of contamination of drinking water is the leaching of lead from water distribution pipes and household 4 plumbing by corrosive water.£ Treatment methods now being used in some communities have reduced lead concentrations below the current MCL of .05 mg/L. If a new, lower standard is adopted, though, additional treatment may be required even in communites currently undertaking treatment.

As part of the regulatory process, the agency has examined the costs and benefits of water treatment for the country as a wh,:,le. To corroborate these analyses and examine the costs and benefits in a more systematic fashion, the Office of Policy

Analysis has undertaken case studies of several cities. As a part of this effort, I was asked by that office to perform a cost/benefit analysis for Boston, a city which has highly corrosiv• water and whose water distribution system is representative of older urban areas.

~nd potential treatment methods, this study will look at Boston's situation and its history of corrosion control. The analysis itself will begin with an examination of the additional treatment likely to be employed in Boston and its cost. It will then turn

I;,:, t ht: b o:tH: fi t s • Because of data and epidemiological constraints, the benef~ts analysis is limited to avoided costs associat•d with neurological damago: in children, hyperten~ion and related cardiovascular disease in adult white malo:s, aged 40-5'3, and reduced materials damage. This paper will conclude with a

1 U.S. Etivi r ,:,nment al Pr ,:,t e,: ti ,:,ti Aget,,: y, "F.:egul at i 1'1Q C,:,r r ,:,si ve Watt:-r", Offi,:e ,:,f F"lat1t1it1g at1d EvalLtati,:,ti, April 1'381, p.2. 5 comparison and discussion of the costs and benefits and

BACKGROUND

The problem at hand stems from the unfortunate coincidence of highly corrosive water and the us• of piping materials containing substances which, when leached, contaminate drinking

,,.,at •r. While this analysis is primarly concerned with the health

impacts of lead and general materials damage, corrosive waters may also contribute other substances including, cadmium, asbestos, iron, and copper to drinking water.

Despite our longstanding knowledge that lead is harmful to human health, several characteristics have made it popular as a material for water piping. It is· easy to form, cut and join.

It is also durable and resistant to subsidence and frost. 2

Because of its durability, many lead pipes installed in the early part of this century are still in use. In addition, a number of building codes still allow its use for joining conveyanc~ pipes. 3

L•ad can be found at many points between the water source and the consumers' tap. Although utilities once used lead lined water mains, many of these have been replaced. Cur r et, ·t l y , a major source ~f concern is the use of lead in sevice lines, the

2 American Water Works Association (AWWA), Internal Corrossion of Water Distribution Systems: Cooperative R•search Report, 1985, p. 214.

3 U.S. EF'A, "Statemer,t c,f Basis arid F'Ltrp,:,se- f,:,r Amer1dme-r1ts t,:, t~,e­ Na·t i ,:,r,al Ir,t e-r i m Primary Drinking Wat er Regt.tl .:il\t i ,:iris", Off i ,: e- ,:, f Drit1kit1g ~,later, 1·:;180, p. 48. E, piping which connects a building's plumbing with the main. While many of these pipes have been replaced over the years, a large number remain in use. Lead can also be found in goosenecks, caulking, gaskets, solder and plumbing fixtures. 4

Although solder and plumbing fixtures constitute proportionally a small amount of interior surface area, studies have shown that these sources can contribute a significant proportion of the lead found in drinking water. A British study in 1977 found lead levels in houses without l~ad pipes as high as those found in lead plumbed houses. 5 These results were confirmed by a second study by Lyon and Lenihan who collected water samples from a modern office building with lead based solder but no lead plumbing. Forty four percent of the samples exceeded .1 mg/L.• Studies also show that while lead levels decrease repidly with the age of the soldered joints, contaminaticn will persist for many years. 7 Laboratory studies have corroborated th~se findings and have also demonstrated analogous problems for brass and bronze plumbing fixtures.a

These results are a cause of concern. While the incidence of lead in service lines, mains, and internal plumbing has

4 Karalekas, P. et al., "Lead and Other Trace Metals in Drinking Water it, the B,:,st,:it, Metr,:,p,:,litat, Area", Pr,:11:e.,..dit1gs AWWA •351- ... Annual Conferenc'=", Minneapolis, Minnesota,June 9-12, 1975, p. 7.

5 AWWA, Op. cit., p.215

6 Ibid., p.216

7 MLtrrell, N,:,rrrran, "Irt,pa,:t ,:,f Me-talli,: S,:,lders ,:,ti Water Quality", Sp.,..cialty Conference on Environmental Engineering, EE Division, ASCE/Boston, MA, July 1-5, 1985.

8 AWWA, Og. cit., p.222 7

declined, the vast proportion of building plumbing systems still

use lead-based solder. This fact suggests that lead may be a problem wherever water is corrosive, not only in older urban

areas where lead plumbing and service lines can still be found.

In fact, newly soldered joints will contribute lead even in areas where water supplies are not corrosive.

The other part of the problem is corrosive drinking water supplies. A large number of water utilities su~ply corrosive tr, ,...,~:t,=r• It, tw,:, stLldies, ,:,tie by Millet·!;,.;. at1d the ,:,ther by t~,e

Midwest Research Institute CMRI), over two-thirds of the utilities sampled reported distributing water that is either moderately or highly corrosive. Millette's study examined 130 utilities serving more than 40 million people. The MF.: I samp 1 e included 388 utilities serving more than 103 million people.g

A number of factors contribute to the corrosive nature of water •. Among them are low pH, alkalinity, hardness, di ss,:,l "ed

,:,:,;ygen, diss,::ilved s,:,lidsv vel,:i,:ity 'cmd temperature. 1 •->

Contaminant levels are also greatly affected by the length of time that the water is in contact with sections of pipe. When water has been standing in the service line and internal plumbing, it tends to produce higher levels of contaminants. 11

An important characteristic that affects corrosivity is the hardness of water, that is, the concentration of calcium

·3 U.S. EPA, "Statemet1t ,:,f Basis at1d F'urp,:,se f,:,r Amet1dmet1ts t,:, tl-1€6 National Int•rim Primary Drinking Water Regulations'', pp.37-40.

10 U.S. EPA, 11 R€-gulatit1g C,:,rr,:,:;ive Watt:"r", p.3.

1,1 U.S. EPA, "Stat,?-ment ,:,f Ba.sis at1d F"Ltrp,:,st:" f,:,r Amet1dmet1ts t,:, 11 th* Nati c,t1al Interim P1rimary Drit1kit1g Wate-r F.:egulati,:,t1s , p.31. 8 carbonate CCaCO~). In hard waters, where high concentrations of

CaC03 are present, a protective skin forms along th~ inner walls of pipes.i 2 Soft water tends to be corrosive because the CaCQ3 conc•ntration is low which inhibits the formation of the protective film. i3 Before discussing the various methods currently available to lower lead levels, I would like to briefly discuss the kinds of health effects associated with exposure to lead.

Lead has long been implicated for its damage to the brain and the central nervous system. At low-level exposure, the concern is especially great for children who retain proportionally greater amounts of lead and who are also going through critical stages in the development of the brain and cognitive abiliti~s, making them even more vulnerable. 1 ~ Studies show that exposure to lead may cause anemia and renal damage; at high levels, it can result in encephalopathy and death. 1 ~ Researchers have also demonstrated a negative relationship between blood lead levels and IQ. In addition, epidemiological studies have demonstrated a relationship between elevated lead levels in pregnant women and low-level fetal malformations. Elevated blood lead levels in pregnant women also may lead to still births and miscarriages. 1 •

1~ Ibid., pp.31-32

13 Ibid., p.35

14 U.S. EF'A. "F.:1:-gulating C,:,rr,:,sive Watei,.", 1"381,p.6.

15 Ibid., p.6.

16 U.S. EPA, Costs and Benefit= of Reducing Lead in Gasoline, 1985, chapter IV. Studies from the mid-1970's have suggested ~hat there is a

relationship between soft water and cardiovascular disease (CVD).

While some evidence indicates that the mechanism for this is soft

water's defi,:ieti,:y ir, magt1esiL1.m and ,:alcium, substar,,:es that may

have a protectiv& effect on blood pressure, leachate contaminants

common is soft water such as lead can increase blood pressure,

and with it, the incidence of CVD. 17 In fact, epidemiologists

have uncovered a very strong relationship between blood lead

levels and blood pressure. According to a study by Pirkle et

al. , a 37¾ drop in blood lead was associated with a 17.5¾

1 reduction in cases of hypertensiorr and lower rate of cvo. •

Short of the wholesale replacement of distribution pipes,

water utilities can lower lead concentrations by controlling

Control techniques currently used include pH

adjustment, hardening, and the addition of silicates or

p~1,:,sphates. The choice of method depends upon both the

characteristics of the water and the types of materials used in

the distribution system. 1 •

Adjusting pH is a widely used method to treat corrosive

water. Most utilites that adjust pH use lime. In additi,:,t1 t,:,

raising pH, lime increases alkalinity and the hardness of water.

The high concentration of calcium promotes the formation of the

17 U.S. EF'A, "Statement ,:if Basis arid Purp,:,s€' f,:,r Amendmet1ts tc, the Nat io::itial Ir,t er i m F'r i mary Dr i r1k i t1g Wat€'r F.:egL1l at i ,::it11;; 11, p. •=+5.

18 Pirkle t!>t. al., "Th€' F.:€'lati,:1t1ship Betw€'et1 Bl,:11:,d Le-ad Lev€'1s arid Bl,::i,:,d Pr-=ssure atid its Cardi,:,vas,:ular 1:.:isk Impli,:ati,:,r1s", American Journal of Epidemeology, 1985, 121:246

1 '3 U.S. EF'A, 11 State-r.-1€"t1t ,:,f Basis and F'L1rp,:,s€' f,:,r Amt?-t1dmet1ts t,:, thE- Na-t;i,:,nal It,t€'rim Primary Drit1kit1g Wat€:'r F.'.;;..gulati,:,r1s", p.26. 10

¢rotective CaC03 film. With respect to ~he dissolution of lead, pH adjustment facilitates the formation of an insoluble lead

precipitate that adheres to the inner walls of pipes and prevents

metallic lead from being further attacked. 20 The American Water

Works Association

lead dissolution approaches a minimum when pH approaches 9. 21

Corrosion can also be controlled with the use of corrosion

inhibitors (e.g. zinc orthophosphate), chemical additives which help form a protective film. While phosphate treatments have

been used for over a decade, intensive research has taken place

in only the last several years. These studies indicate that

treatment with zinc ·orthophosphate is effective, but only within certain pH and alkalinity constraints. 22

Aside from lowering lead concentrations, corrosion control has lowered the levels of other contaminants and reduced damage

to water pipes and other distribution sy~tem components~ As a result, any analysi• of the costs and benefits of corrosion control must not only consider the benefits from reduced exposure to lead but also from the reduced exposure to other contaminants, avoided materials damage and improved eesthetics.

Cadmium, another material found in distribution pipes and

leached by corrosive water is also linked to hypertension. It is found at high levels in hypertensives and has induced 21) Kar al ekas, "Alt en,at i ve !'1eth,:,ds f,:,r C,:,ntn:il 1 i ng the Corrossion of Lead Pipe'', Journal of the New England Water Works Association, June 1978, p.2.

21 AWWA, op. cit., p.243

Ibid., pp.246-260. 1 1

hypertension in animal studies. 23 Cadmium also may cause

irreversible renal damage. 24 Concern also has been expressed

over the use of asbestos lined concrete pipes. While asbestos is

a known carcinogen when inhaled, its health effects when ingested

have t11:,t been wel 1 defi tied. 29 Because asbest,:,s pipe is ti,:,t used

in Boston and cadmium levels are acceptably low, contamination

from these substances was not examined in this analysis.

In addition to the contaminants regulated by primary

standards, there are substances subject to secondary standards

that also are leached from pipes. These include iron and copper.

EPA regulates copper on the basis of smell and taste.

Concentrations greater than the 1 mg/L standard may also stain

sinks and porcelain. Likewise, iron is also regulated on the

basis of aesthetic considerations.a&

Finally, corrosive water damages pipes and other components

with which water comes in contact. The process of leaching

minerals causes more rapid interior degradation of both

distribution pipes and privately owned plumbing pipes, making

them mere susceptible to leakage and rupture. 27 Leakage from distribution systems may be substantial, sometimes accounting for

23 U.S. EF'A, 11 F.:egulatit1g c,::irr,:,sive Wa-1:;er", p.8.

24 U.S. EF'A, 11 Statemer1t ,:if Basis and Purpr.:,se- f,:,r Ame-ndmt\'t1ts t,:, ·the Nati ,:,tial Int•rim Primary Drinking Wat•r Regulations", p. 44.

25 U.S. EPA, "Regulating C,::,rr,;:;sive, Wate-r", p.8.

26 Bureau of Water Works, Portland,Oregon, Internal Corrosion Mitigation Studv: ~inal Report, 1982, p.5-12.

27 U.S. EPt,, 11 F.:egL1latit1g C,:,rr,:,sivt?- Wat.,:r", p. 1;'3. 12

33¾ of the water supply. 2 • In addition, corrosion promotes

tuberculation, a process in which leached materials build up on

the inner walls of pipes reducing their carrying capacity. 29

This deterioration leads to reduced flow and the need to increase

pumping on the part of the utility. Various studies have shewn

that the costs of corrosion damage are substantial.

BOSTON: ITS WATER AND CURRENT TREATMENT

The water supplied to Boston by the Massachusetts Water

Resources Authority CMWRA), the regional 'wholesale' water

utility, is among the most corrosive in the country. It is

relatively acidic with a pH of 6.7, and soft with hardness

measured at 12 mg CaC03/L. Alkalinity is _low as well. 30 Boston

is also a city with an old distribution system and housing stock

and has a significant number of lead services still in existence.

Although local officials have long recognized the dangers

inherent in the use cf lead pipe for supplying potable water and

have commenced the systematic replacement of lead service~,

property owners are responsible for that portion of the service

line that runs from the property line to the structure. As a

result, staff engineers at the Boston Water and Sewer Commission

28 U.S. EF'A, "Stater.-1et1t ,:,f Basis and F'urp,:,se- f,:,r Amet1dmet1ts t :, the Nati ,:,t,al Ititerirn Primary Drit1kit1g Water F.:egulati,:,n-:s", p.27.

29 U.S. EPA, Regul_ating Corrosive Water, p.13.

30 Karal-=:-h:as et. ::ll., "C,:,t1trc,l ,:,f Lead, C,:,pper, ,~t,d Ir,:,1·1 F'ip1:­ Cort0sion'', Journal of the American Water Works Association, 1 "383, p • "=J3. 13

CBWSC) estimate that 44¾ of their residential customers still

have lead services. 31

Some studies, however, indicate that the lead problem could

be even more widespread. The most comprehensive tap water

sampling in Boston indicated that 70% of the household showed

evidence of lead dissolution. 32 Since only half the houses in

the survey had lead services, we must assume that lead is being

leached out from other sources. F'r,:,m t~,e previ,:,us dis,:uss..i,:it1

concerning lead solder's contribution to lead levels in drinking

water, it seems likely that solder accounts for mu~h of the

,:,:,titami nati ,:,n. Goosenecks and caulking also contribute lead.

Because of the widespread existence of lead in piping

material, tap water was extensively monitored in th• mid 1970's.

This study revealed a large proportion of samples in excess of

the .05 mg/L MCL. 33 In addition, a study by Worth et al. showed

a statistically sianificant relationship between lead in tap

water and blood lead in children. 3 • In response to this

situation, the Metropolitan District Commission CMDC), the former

water supply agency, began treating the region's water.

31 BWSC staff engineers, January, 1986.

32 Karalekas, uL~ad and Other Trace Metals in Drinking Water in the 8,:,st 1: 1t1 Metr,:,p,:,litat, A1~ea", p. 7. -:, 33 Ibid., p. 1w •

34 L•J,:,rth et. al. "Th

At first, the MDC used zinc orthphosphate but was unsuccessful. 38 (It now seems that the failure of orthophosphate

3 was due to the low pH of Boston's water. •) Beginning in 1977, the MDC began pH adjustment using sodium hydoxide CNaOH), a control technique that has proven to be extremely effective. The

MDC chose NaOH over lime because its consultant, Metcalf and

Eddy, estimated that capital and operating and maintenance costs were significantly lower for NaOH treatment. 37 In addition, sodium levels in MDC water were sufficiently low that adding sodium would not create a health problem. 39

Monitoring performed by EPA's Region I office from 1976 to

1981 indicated that lead, iron and copper levels dropped ' significantly. 3 • More importantly, lead levels in most water samples fell below .05 mg/L. Most samples, however, had levels which remain above the contemplated .01 mg/L MCL. with the proposed standard will require corrosion con~rol to

further reduce lead 1evels in Boston's drinking water.

35 Karalekas, "C,:,ntr,:,l ,:,f Le-ad, C,:,pper, and Ir,:,r, Pipe C,:,rr,:,si,:,n", p. '34.

36 Discussion with Peter Karalekas and studies reported in the AWWA corrosion control study, December, 1985.

37 Discussion with Peter Karalekas, April,1986.

38 Ibid.

3,;i Karal•kas, "C,:,t1tr,:,l ,:,f Lead, C,:,ppt?r, and !1·,:,n Pipe- C,:,r1•,:,si,:,r1", p. '33. 15

PROPOSED TREATMENT AND ITS COST

The primary objective of additional corrosion control

is to reduce lead conc~ntrations in order to comply with the proposed .01 mg/L standard. Before deciding upon which treatment meth,:,ds ,...,,:,L1ld bt? appr,:,priate, we must first establish t~,e criteria by which we define compliance with the standard. Bas .... d on recent epidt?miological studies and technical feasibility, this analysis proposes the following compliance crit~ria. ~irst~ the

MCL should be based on the standing grab sample, that is, the sample which is taken immediately after turning on the cold water tap and the one which is statistically the best predictor of blood lead levels. Second, compliance should be based on tap water collected from a sample of worst case households, those which have new lead soldered joints, lead servi~es, or other evidence of lead pipes. Because it is impossible to guarentee that every household could me~t the .01 mg/L standard even using state of the art corrosion control techniques, we should use the mean concentration generated by this sampl•.

We must also be cognizant of a number of uncertainties concerning the effectiveness of corrosion control techniques. A particµlar form of treatment might produce excellent results in one system, and yet, performance in another may prove less effective. In addition, some methods havt? been extensively t~sted ih the I~b but not in the field. Th• treatment methods I will discuss are endorsed by the new AWWA manual on corrosion

,: ,:,t1 tr ,:, 1 • Th~y also have b~en sugg~st~d as th~ likely 1 E,

alternativ~s by engineers at EPA Region I and the MWRA and have

prov•n ~ff•ctive in other New England cities where the

characteristics of the water -and the distribution systems are

similar. Nevertheless, our estimates of expected reductions in

lead ,:,:,n,:entrati,:,ns are based ,:,rily ,:,ti e-du,:ated judgemen,ts.

After reviewing the current research literature on corrosion

control, and discussing the Boston situation with EPA and MWRA

personnel, it is likely that additional treatment would consist

of two stages. 4 ~ ~irst, MWRA would further raise pH· by increasing

th• NaOH concentration. When the MDC first began NaOH treatment,

th• state of the art suggested that pH should be elevated to a

range of 7 to 8 or higher. Currently the pH of tr~ated water is

9.5. Recent findings, however, indicate that corrosion control

will be even more effective when pH is raised to 9.

In addition, it is likely that th~ utility would take action

to achieve c,:,tisi stent pH 1 evel s thr,:,ugtrc.,Ltt the ,:,:,mpl e-:,; web ,:, f t~,e

MWRA distribution system •• This would necessitate the installation of several additional pumping stations to even out

the concentration of NaOH throughout the delivery system.

While additional and better controlled pH adjustment

should reduce corro•ion, the use of a corrosion inhibitor will

probably be necessary. Zinc orthophosphate is the likely

alternative. Although this method was used unsuccessfully before

in Boston, it now seems that pH was too low. W~iet, Llsed i ti a

40 Discussions with Peter Karalekas ~f EPA and Guy Foss of the MWRA. Mr. Foss was reluctant to make any judgments but indicated that higher pH and zinc orthophosphate was a likely scenario, Doi-•: ember, 1 •~95. 17 high~r pH environment, corrosion control experts believe that orthophosphate will be very effective. In addition, the MWRA does not exp~ct a repeat of the phosphate induced algae growth experienced in 1976 in tne smaller storage reservoirs because they are no longer used, a l~vel that would comply with the contemplated MCL. • 1

Like Boston, Bridgeport, Connecticut also has low pH water supplies and widespread use of lead in the distribution system.

Using lime to raise pH and zinc orthophosphate, water quality authorities in Bridgeport have successfully reduced corrossion.

Samples for lead indicate a mean concentration of .007 mg/L overall and .01 mg/L for the standing grab sample.•=

The techno~ogies associated with corrosion control are relatively simple and the costs are calculable. M,:,st treatment methods employ processes similar to those used for chlorination and fluoridation. The chemicals are held in storage tanks, mixed with water, and then pumped into the distribution system. Thus, the capital component includes the costs of storage tanks, mixing equipment, pumps and installation labor. Operating and maintenance costs CO&M: labor, energy, repairs) are moderate.

Usually they amount to approximately 20¾ of the capital outlays.

The expenditure for chemicals constitutes the largest component for any kind of corrosion control treatment with lime being the

,:he,apest atid z i ti,: ,:,r th,:,ph,:,sphat e am,::itig the m,:,st e:,;petisiN-e-..

41 Guy ~oss, MWRA.

42 Karalekas, "Alternative Methods for Controlling the Corrossion 11 ,:,f L>?ad F"ipe- , pp.7-•";J. 18

The pH adjustment phase of the additional treatment can be

broken down into two components: additional pumping stations and an increased chemical feed rate. Because of the low buffering

capacity of Boston's water, only a small increment to the feed rate will be necessary to raise the pH to the desired level.

Both the EPA and MWRA estimate that chemical costs would increase by approximately 10%. Because NaOH is a by-product of automobil~ manufacturing and output varies directly with automobile output,

its price has fluctuated dramatically. Based on the average

price for the last five years, the cost of NaOH is S.~4 per mg per million gallons Cmg/MG, 1995$). 43 At that price, raising chemical costs by 10¾ would amount to an annual incremental increase of $85,000 (1985 $; all costs and benefits in this paper are in 1985 $ unless otherwise specified), assuming an average

flow of 310 million gallons per day CMGD).

The cost of an additional two 50 MGD NaOH feed systems including installati6n is $150,000. 44 Amortized at 5¾ over 20 years, the resulting annual capital cost is $12,000. Operating and maintenance costs are $30,000 (20% of $150,000). Therefore phase I would cost approximately $127,000 per year. (See TABLE I for a summary of treatment costs.)

Although the per unit chemical cost for zinc orthophosphate is much higher, its fe•d rate tends to be low. While the literature cites several different cost estimates, this study

43 NaOH cost figures supplied by th~ MWRA.

44 U.S. EPA, ''Corrosion Manual for Internal Cor~osion of Water Distribution Systems'', Prepared for the Office of Drinkinq Water by Environmental Science and Engineering, Inc., 1984, p.1~3. TABLE I

TF.:E,~ TMENT COSTS

PARAMETERS averag,;.o fl,::,w 3:LO itite-rt?st rat~ S¾ payba,: k pe-r i ,:,d 20 CF.:F 0.08024

,: ,:,s ·I; ( r.-,g / M13) $0.54 feed ratt? Cmg/L) 1.4

,::he-mi,: al s $85,541 Qa,.M $30,000 ,:ap ital ·:~12, !)36 t,:,tal $127,578

Zinc Orthophosphate-

EPA (A) ,:,:,s·t (mg/1"113) $5.44 h•'='d ratt? (n·,g/L:>

,: h,a,mi ,: al s $1,231,072 081.M $:34, ()()() ,:ap i ·tal $:L3,641 t,:,tal $1,278,713 EPA (8) ,:i::,st (mg/MG) $5.44 feed ratt? (mg/L) 3

,: ~1,a,m i ,: a 1 s :i:-1., 846,608 OM1 ·$34,000 ,:apital ,H3, 641 t ,.:,t al <1; 1 , 8"::14, 24':'

T·e,:h Pr"oduo::ts cost (mg(Zn)/MG) 921.00 f•'='d rate Cmg(Zn)/L) 0.6

=$ :L , =+25, 6'30 -~3<),, c··,C=C capit,:it:L ~s :t 2 ,, -:::., ~3 i,:; I; ,.:.-i; -ol.i TABLE I (cont.>

AWWA co5t Cmg(Zn)/MG) S19.20 f~~d rat~ Cmg(Znl/L) 0.5 chemicals $1r086,240 O&M $30,000 capital $12,036 total Si,128,276

EPA CB) 12,021~8~7 TECH PROD 11,595,304 EPA CA) $1,406,291 AWWA $1,255,664 1 "::I

uses the figures provided by Technical Products, a chemical firm which manufactures orthophosphate and supplies the feed systems.

Assuming a flow of 310 MGD and a feed rate of .6 mg/L, the annual chemical cost is $1,425,000 based on a unit cost of $21 per mg of

zinc per million gallons CmgCZn)/MG).

The cost of the zinc orthophosphate feed systems are less than those for NaOH syst~ms with a 100 MGD plant costing $30,000.

In order to meet peak demand, the MWRA would require a total of

500 MGD capacity costing $150,000 or $12,000 on an annual basis.

Bas.ed ,:,r, th•s• ,:api·l;al r1:>q1.1il'·em•r1ts, t~,& 1:.6:,;pe,:ted am,ual O&M costs are $30,000. The total annual cost for zinc orthophosphate treatment is therefor• $1,468,000. The combined annual cost of pH adjustment and orthophosphate is approximately $1,600,000.

While this cost figure represents the most likely case, th•re ~~• a number of factors which could lower treatment costs.

~irst, .dditional pH adjustm~nt alone could be sufficient.

Secondly, optimal_ pH for zinc orthophosphate could require a lowl:."r NaOH fl:."ed rate and with it, lower NaOH chemical costs. rinally, these costs are based upon MWRA's projected demand of

310 MGD, a forecast which disr~gards price responses Ci.e. if water rates increase, consumers may reduce their demand).

Because of the capital requirements of the Boston Harbor cleanup program, combin•d water and sewer rates are expected to double in r-eal te-rms ,::iver t~,. ,:,:,ming de,:ade wit~, prc,bable- re-du,:·l;L,:,t,s it, demand. Sine~ corrosion control treatment costs are heavily depe~dent on flow, price induced water conservation would lower HEALTH BENE~ITS

Despite the fact that the health effects from exposure to

lead are relatively well understood, monetizing health benefits

is complicated and requires a number of stages and several kinds

of data. In graphic terms, the process is a chain which starts

with a change in contamin~nt concentration producing a change in

exposure which produces a change in body lead burdeh, which

finally results in observable physiological and neurological

changes. By reducing contaminant concentrations, we can reduce

the adverse health impacts of exposure to lead and avoid the

various costs associated with them.

The first requirement of this stage of the analysis,

~herefore, was data for a change in water lead concentration. Based on the definition of compliance established earli•r in this

analysis and the Regi~n I water lead data, we estimated the

reduction in lead concentrations. The mean concentration for the standing grab sample was .032 mg/L. This value is based on the

cross sectional data peeled for the last five months in which

samples were taken. During this time lead levels seemed to

stabilize following the commencement of NaOH treatment. Since

compliance by definition requires that we lower the mean

concentration to .01 mg/L, lead levels would drop by .022 mg/Lon average. Assuming proportional reductions, the decline in lead

concentrations ranges from .05 mg/L to .0055 mg/~ for the hou$eholds in the sample. 21

Be,:au·se the sample size is s,:, small, it is like,ly t~1at this

reduction in water lead in not representative of tap water throµghout Boston. In fact, we know that the houses in the study

were ~elected because they had lead services, a characteristic true of only 44¼ of residential consumers. The standing draw

data used here, though, is the water that is in contact with internal plumbing. Because of this fact, the fact that 70% of the households surveyed (only 50¼ of which had lead service) in

1975 showed evidence of lead dissolution, and the studies which show that lead solder contributes significant amounts of lead to drinking water, it is likely th•t the data are representative of a much larger portion of the Boston housing stock. the most likely case, this analysis uses a figure of S7X Cthe average of 44¼ and 70¼) to calculate the number of houselolds effected by further reductions in drinking water lead levels.

Second, the analysis required blood lead data for the r->l-1...-vat,t segme-r,ts ,:,f the p,:,p1.1la.ti,:,n, d1ildre-r, arid adult 1..Jhi·te­ males, aged 40-59. bl ,:u:,d 1 ead • This data, however, were- not useful since- it was collected and organized for a completely different purpose.

Consequently, this analysis employed data from the National

Health and Nutritional Examination Sutvey CNHANES II). By using national data, however, we are making the assumption that the­

B.,:,st,:,n p,:,pulati,:,r, gr,:,ups ~,ave the same bl,:11:,d lead distrib1.1ti,:,r, as the nation as a whole. But because an urban population is likely to have- blood l~ad l~v~ls high~r than the national average, th~ analytical r~sults will be biased. 22 data also had to be adjusted for changes in gasorine lead content so that we can isolate the impact of a change in water lead

Thirdly, we had to have population estimates for the 1985-

1995 period. The population projections employed in this study are based on the Census Bureau's estimates and reveal several patterns which are significant to this analyis. 4 e Although

Boston's population had declined for most of the post-war era, this trend reversed in this decade. Population incfeases are most pronounced for the two groups studied in this·analysis. The number of young children is expected to grow rapidly as the baby- boomers start to have children. In addition, the ranks of adult whit~ males, aged 40-S9 will swell over the next decade as the baby-boomers age.

Finally, when estimating health benefits, this analysis assumes that the proposed MCL will not be implemented until at least 1988. Thus, the calculations reflect benefits starting in that year and extending out to 1992.

CHILDREN'S BENEFITS

In order to calculate the health benefits for children, resulting froM reduced exposure to lead in drinking wate~, the analysis used a methodolog~ similar to that employed in EPA's

45 Bost,:,n F.:ed€-vel,:1pmet1t Aut~1,:,rity, "P,:,pulati,:,n· F·,~,:,j€:'•:ti·,:,ns f,:,r Boston and for Boston City Hospital Neighborhoods--by Race,Ethnicit¥, Age, Income, and Pverty Status--to the Year 2000, Research Department, 1385, pp.41-44. 23

study of lead in gasoline. Th• derivation of avoid•d costs is

bas•d on the number of childr•n ·whose blood l•ad l•v•l falls

b•low 25 µg/dl as th• result of r•duced l•ad concentrations in

drinking water. W• can estimate this numb•r by taking th• diff•r•nc• b•tween th• numb•r of childr•n with blood l•ad )25

µg/dl at current lead l•v•ls in drinking water and th• predict•d numb•r of childr•n with blood l•ad >25 µg/dl at lowered l•ad

levels in drinking wat•r.

In th• first step, children wer• brok•n into two racial groups, black and non-black, and into sev•n two-y•ar age groups

starting with six months to the second birthday and ending with tw&lv• to thirt••n years of age.~• MINITAS, a statistical software packag~, was then used to generate a distribution of blood lead levels for each ag&/rac• group based on the parameters of th• NHANES II data, controllin~ for gasoline lead l•v•ls •• In order to estimate th• change in th• distribution of blood lead for •ach group, we had to relate r•duced exposure levels as measured by lower lead levels in drinking water to blood lead levels. This analysis uses a 1993 study by Ryu which establish•d th• relationship between water lead and blood lead in childr•n .as d•pict•d in th• following equation:

PbB =a+ .12PbW whe-re: PbB = blood l•ad in µg/dl; .a = the ,:,:,r1st-a~1·t plus ,:,·t~ier ~.1\H1)•:,graphi,: and other environmental fact0rs; and PbW = wat•r l•ad in µg/L (standing grab sai.wpl &) •

46 Children Were categorized in this way because blood lead distributions are heavily dependent on age and race. 24

Ba$ed on this regression equation and the data on water lead reduction, a child in Boston is subject to a po~sible reduction in blood lead ranging from .66 µg/dl to 6.04 µg/dl depending on the reduction in water lead in his/her household.

In the final step, a BASIC program used the initial blood lead distributions and epidemiological relationship above to simulate the change in the blood lead distribution for each age/race group. Using the before and after distributions, the program calculated the probability that a child's blood lead would drop below 25 µg/dl. By applying that probability to the number of children in an age/race group and repeating the process for all 14 groups, this procedure estimated that additional tr•atment of Boston water will reduce the number of lead-poisoned children by 87 in 1988 Csee Table II).

In terms of medical and compensatory education expenses, EPA believes that $3,900 can be saved for each avioded case of lead poisoning: Sl,100 for medical treatment and $2,800 for education.

Th•r•fore w• can expect a total of $340,000 in benefits for children in 1989, rising to $349,000 in 1992.

This estimate, however, merits closer scrutiny. Several factors suggest that it is somewhat conservative. ~irst, the

NHANES II data probably underestimates blood lead levels for

Boston children. Urban children are exposed to higher amounts of lead (automobile exhaust, lead paint, dust, industrial sources) and are therefore more likely to have higher blood lead levels.

By using national data, we start out with proportionallj fewer individuals with blood l•ad >25 µg/dl thus reducing the TABLE II

CHILDREN'S BENEFITS

~NETERS -:ii t p~r ~api ta S3 1 900 . o:!lt 1ppl 11:ib l-: S4X BLACKj

=-::e ~ CHANGE NUNBER BENEf'IiS BENEFITS BENEFITS BENEFITS B£NEFiiS Pi1P _;:;GuP ?DP 25 !laidl CHILDREN 1'388 POP 1989 POP 1990 ?OP 1991 • J, 1·~·~·'• •4-

;•

;_:il)TAL 65,2 S254,439 $256,432 $258,687 $260,294 $261,477

NON·lLACKS

.:.:lE 7. CHANGE NUHBER BENEf'ITS BENEFITS BENEF'ITS BENEFITS BENEFITS :if.CUP POP 25 uq/dl CHILDREN 1998 POP 1989 POP 1990 POP 1991 POP 1992

)·<2yr 7277 0, 135X 5,3 $20,689 . 7371 s20, 0356 7466 $21,227 7475 s21, 252 7484 $21,278 2-"3 9734 0,145% 7.6 S29,72S 9828 $30,012 9960 S30,415 '3970 S30,445 9979 $30,473 1-5 'HlO 0,120t 5,9 $23,023 "3167 $23,167 '3223 $23,308 '3308 S23,523 '3356 $23,644 ~-7 8486 0,06S% 3,0 511 1616 8505 $11 1642 8514 SU,655 8647 Ul,837 a1s1 $11, '387 '3•'3 7862 0.000% o.o so 7844 so 77~6 so 798S SO 8146 $l) :0-1 l 7239 0.000% o.o so 7182 so 7088 $0 7308 $0 7529 so ;,;•iJ "6615 t),0001. o.o so 6521 so 6379 $1) 66S3 so 6930 $1) t:iOTAL 21.8 SSS,053 sa5, 778 $86,605 ·&a7 ,os0 567,383

.I~ . 87.Q $339,492 .l342 1.210 $345,292 $347,351 $348,959

25

probablility that a blood lead level will fall below the CDC

blood lead level criteria for lead poisoning.

Secondly, the methodology employed in this analysis

implicitly assumes that water lead and blood lead are independent of each other. This ignor•s the fact that higher blood lead

levels are the result of greater exposure to lead, and probably, people who have higher blood lead levels are more likely to have

higher lead lead levels in their drinking water. A more realistic methodology would produce a higher probability that a child falls below 25 µg/dl because it would assign the larger reductions in lead exposure from drinking water to high lead children when simulating the new blood lead distribution. Under the independence assumption, however, high lead individuals fate the same reductions as low lead individuals, reducing the probablity of falling below 25 µg and biasing down thP benefit estimates.

Finally, and most importantly, the monetary estimate comprises only avodied medical costs associated with the treatment of lead poisoning and avoided expenses for compensatory education. A number of benefit categories have been excluded from the analysis because the relationship between blood lead levels and certain health effects have yet to be precisely specified. S~ch benefits include avoided costs as~ociated with renal damage, a very serious effect of lead exposure, increased risk of anemia, vitamin D deficiency, and permanent nerve damage.

The analysis also includes benefits that are exceeding difficult to quantify such as avoided pain and suffering associated with 26 medical care and reduced quality of life resulting from the permanent •ffects of damaged cognitive development.

HEALTH BENEFITS -- ADULT WHITE MALES, AGES 40-59

Although the public health community has long known about the link between high lead exposure and elevated blood pressure, it is only recently that research has uncovered effects at low

Prior to the analysis of the NHANES II data by CDC and EPA personnel, a number of investigations reported a statistically significant relationship between low to moderate blood lead tevels and blood pressure in males. In addition to epidemiological analyses, animal studies also demonstrate this link.•7 The research also indicates possible causal pathways by which lead acts ori the cardiovascular system. renal chan~es and inhibited uptake of calcium, an element which suppresses blood pressure. Besides dem,::instrat i ng a str,:,ng a.nd significant relationship between blood lead levels and blood pressure, the analysis of the NHANES II data showed that there was no threshold level of exposure. In other words, there are blood pressure effects at any bl09d lead level down to zero. 4 •

Because of lead's direct contribution to hypertension, the toxin is also associated with cardiovascular disease (CVD) resulting from elevated blood pressure. Two extensive studies,

47 U.S. EPA, Costs and Benefits of Reducing L•ad in Gasolint, pp.V..:4 - V-5.

48 Ibid., pp.V-6 - V-15. 27

the Framingham and the Pooling Project, have assessed the risk of - .- ? -r CVD (strokes, myocardial infarctions, and deaths resulting from

all forms of CVD) based on several important variables, including

blood pressure, serum cholesterol, and smoking. The

corresponding risk regression equations show a very strong

relationship between CVD and blood pressure.~9 Sinca blood

pressure seems to increase with blood lead, so. too will the incidence of CVD.

To calculate reduced cases of hypertension and· incidence of

SVD~ we must determine the impact of lower lead exposure from

drinking water on blood lead levels in adult men. While a nL~ber of epidemiological ·studies have been performed which relate water

lead to blood lead in adults, the Pocock study is perhaps the

best. His study is especially relevent to this analysis because

i.t measures effects at lower water lead c~~centrations. His

findings are summarized in the following equation:

PbB =a+ .06PbW where: PbB = blood lead in µg/dl a = constant plus other demographic and environmental factors, and PbW = water lead in µg/L (standing grab smaple).

Using this regression equation and the estimate of reduction in

water lead levels, the mean reduction in blood lead is estimated

to be 1.33 µg/dl C.06 * 22.2 µg/L). Accounting for changes in gasoline lead, blood lead for the avfrage male will decline from 8.25 µg/dl to 6.92 µg/dl in 1988.

49 Ibid., pp.V-29 - V-31. 28

Hyp~rt~nsion

Medical authorities consider adults having diastolic blood pres5ure greater than 90 to be hypertensi~e. Elevated blood pressure dramatically increases the risk of all forms of cardio- vascular disease and thus requires medical attention. Treatment costs for hypertension include visits to a physician, medication, ho~pital stays, and the opportunity costs of lost working days, and when combined, total $250 per year.e0 While, calculating the reduction in cases of hypertension

for adult males employs on a fairly direct method, the process in this analysis is complicated by that fact that the logistic regression developed by Pirkle and Schwartz (1985), which estimates the probability that an adult male will be

hypertensive, performs poorly at b 1 ~od lead levels <10 µg/dl.

Consequently, this analysis does not make a quantitative

While it is likely that there will be a reduction in

cases of hypertension, the number will be small and the omission

will not seriously affect the outcome of the analysis. It is~

however, one of the factors which contribute to a conservative estimate of the benefits.

Cardiovascular Disease

Having previously estimated the reduction in blood lead resulting from reduced lead levels in drinking water, calculating

50 Ibid., p. v-js. 2•3

the reduced incidence of CVD requires that we first estimate the

change in blood pressure, and second, estimate the change in risk of suffering from various forms of CVD.

By using the before and after blood lead values (i.e.

8.25 µg/dl and 6.92 µg/dl) in the NHANES II regression equation

linking blood lead and blood pressure, we calculated the change.

in blood pressure. We then used the before and after blood

pressure values in the CVD risk regression equations in order to estimate the change in probability that an individu~l will suffer

By applying the changes in probability to the adult white male population, we were able to e~timate the avoided

incidence of CVD.CSee table III)

Myocardial Infarctions. In this analysis, the benefit estimate

for Myocardial Infarctions (Mt~) is based on an EPA calculation that includes medical costs Cbuth physician and hospital) as well as lost wages.~1 Although fatal Mis are included, the benefits measure in this category does not account for the value of a lost life. This valuation is treated in a later section.

The estimate of avoided Mis is based on the Pooling

Project's risk assessm~nt in which the probability of suffering a

MI was calculated as a function cf age, smoking behavior, serum cholesterol, and diastolic blood pressure. We derived the change in the probability of suffering an MI from the change in blood pressLtre. Multiplying the probability differential by the population yields the estimate that 2.95 Mis can be avoided in

51 Il::>id., p. V-:38. TABLE I I I

ADULT BENEFITS

PAPAMETERS !::l~rum chole-sterol 218 ~·,+ra9e- age- 50 ~~«•l 1ne blood pb 17.4 bl,:,,:,d pb redu,:ti,:m 1.33 pef~~nt applicable 54¾ AVOIDED MYOCARDIAL INFARCTIONS PARAMETERS avoid~d cost p•r MI $65,600 smoking 0.8

1988 1989 1'390 1991 1992

staYt blood lead, 8.25 S..23 8.2:2 8.20 a. 1•3 end blood lead 6.92 6.'30 6.89 6.97 6.86 stayt blood pressure 81.53 81.51 81.51 81.50 81. 49 ~nd blood pres!\Sure 80.72 80.70 80.70 80.68 80.68 staYt prob MI 0.00658 0.00658 0.00658 0.00658 0.00657 •nd pr,:,b MI 0.00643 0.00643 · o. 00643 0.00642 0.00642 Di f pr,:,b MI 0.00015 0.00015 0.00015 0.00015 0.00015 population 36100 36800 37500 38800 40250 avoided MI$ 2.'35 3.01 3.07 :3,1'3 3. ~31 benefits $193,410 $197,728 $201,682 $209,074 $217,096

AVOIDED STROKES PARAMETERS avoid•d cost per stroke $49,000 smoking 8. 12

1'388 1'38':I 1'3'30 1'3'31 1'3'32

start blood lead 8.25 8.23 8.22 8.20 e. 1'3 end bl ,:,,:,d 1 e-ad 6.·32 6. '30 E,. 8'3 6.87 6.86 ~tart blood pressure 125.66 125.63 125.63 125.61 125.60 end blo,:,d pressure 124. 16 124. 13 124. 12 124. 10 124. 0':i start pr,:,b s,tr ,:,k e. 0.00347 0.00347 0.00347 ,). ,)0347 '). 00:347 .:-nd prob str,:,ke 0.00338 0.00338 0.00338 0.00338 0. (H)33/ Di f pro:ib stri::oke o. 0000'3 0. 0000'3 (l , (H)(!(l 'j o. 0000'3 1). 0000'3 populati,:,n 36i00 36800 375(H) 388(H) 4025(> avo1 ded str,:,kes 1.75 1. 7'3 1.82 t ,8'3 1. '36 ben~fits $114,642 $117,194 $119,534 $123,910 $128,661 TABLE I I I (,:,:int.)

AV• IDED DEATHS

PAF.:AMETERS avoided cost per death $1,000~000 sm,:,ki t1g 2.24

1'388 1'38'3 1'3'30 1 "3'31 1•:i-:i·-::, bl ,:11:,d 8 .-,,:: 8 .-,.-, ·3tart 1 t.'ad • ~..,j • ~...:i 8.:22 8.20 :3. t ·? o:t1d bl ,:i,:,d lead 6. "32 E,,. '30 6.8'3 6. :37 6.86 star r, bl ,:11:,d pr i:-S-;Ltr e 81.5:3 81.51 }3!.51 81. 50 81.4"3 1:'t,d bl,:i,:,d pressure 80.72 80.70 80.70 80.68 80.68 ·5tar t pr,:,b death 0.01314 0.01314 0.01314 0.01313 0.01313 et1d pr,:ob death 0.01282 0.01281 0.01281 0.01281 0.01281 Di f pr,:,b death 0.00032 0.00032 0.00032 0.00032 0.00032 p,:,pul ati c,n 36100 36800 37500 38800 40250 .av,:oi ded Mis 6.28 6 7,:, 6.42 6.55 . , - 7.05 benefits $6,278,613 $6,418,852 $6,547,245 $6,787,263 $7,047,689

TOTAL CVD $6,5861665 $6,733,774 $6,868,462 $7,120,247 $7,393,446 30

1988, increasing to 3.3 in 1992. In monetary terms, we valued an

MI at S65,600, yielding total annual benefits avoided Mis at

S193,000 in 1988, rising to $217,000 in 1992.~2

Strokes are a debilitating form of cardio-vascular

disease that can leave parts of the central nervous system permanently damaged. In estimating the benefits achieved from avoided strokes, the analysis includes only medical costs and

f ,::,r ti-g or, e wag es.

The estimation procedure uses the Framingham study risk regression equatio~ assessing the probability of suffering a stroke based on systolic blood pressure, age, smoking behavior and serum cholesterol. 153 Using the same procedure employed for

Mis, we estim~ted that 1.75 str6kes can be avoided, rising to

1.96 in 1q92; at S49,000 per stroke, this produces benefits of

$115,000 6n 1988 and $129,000 in 1992.~~

In addition to assessing risks for strokes, the

Framingham study estimated risk regression equations for deaths as a function of diastolic blood pressure, smoking, and chol~stli'rol. The study looked at death from all causes associated with blood pressure, not only myocardial infarctions and s;trokes.~15 ~--.. ..;.,,:;. I!:Jid., P• V-38 V-3•:,.

53 Ibid., p. V-31.

54 Ibid., p. V-40.

55 Ibid., p. V-31. 31

Again using th• blood pressure calculations from above, we substituted the blood pressure values into the risk regression equation to calculate the change in .probabi1ity. We then applied the probability differential to the Boston population to e-ti~ate the reduction in expected deaths. This processs yields the estimate that 6.3 deaths can be avoided by reducing water lead levels through additional corrosion treatment, a figure that rises to 7 in 1992.

Monetizing the benefit of saving lives has alwgys been controversial. We used the somewhat conservative estimate that saving a statistical life is worth a million dollars. Thus, the benefits in terms oi lives saved in 1988 is $6,300,000, increasing ~o $7,000,000 in 1992.

Total Benefits for Adult Males

Having calculated the avoided myocardial infarctions, strokes, and deaths, I estimate the total benefit for this age/sex group of the ~opulation to be $6,600,000 in 1988 and

7,400,000 in 1992.

There are several substantive issues, though, that suggest that this benefit measure is somewhat conservative. First, we have excluded non-black men outside the 40-59 age group and as well as blacks of all age groups. Blacks were excluded because the Framingham and Pooling Project studies did not have a sufficiently large sample to develop risk assessments. With respect to the age issue, it is difficult to seperate the effects of age and blood pressure outside this age group. It is likely, however, that exposure to lead will increase the incidence of CVD in these other population groups.

Secondly, the analysis excludes other kinds of health

•ffects of lead exposure such as renal damage. These effects have not been included because medical and epidemiological research has yet to determine precisely the relationship between blood lead and extent of physiological damage. While renal problems can often be treated, medical care is very expensive and often accompanied by adverse emotional impacts. In addition, renal damage affects all age and sex groups. It is therefqre a very important omission from the benefits calculation.

Thirdly, the cardiovascular monetized estimate fails to consider quality of life issues. The benefits of avoided myocardial infarctions and strokes included only medical costs and lost wages. They fail to account for the kind of limits placed on the lives of heart attack survivors. Stroke victims who may suffer from partial paralysis and loss of speech facilities dramatically diminish the quality of life.

On the other hand, the analytic methodology used to estimate

CVD related benefits may lead to an upwardly biased measure. The first bias results from the use of national blood lead data for adult men. Because adult males in Boston probably have higher blood lead levels than their counterparts in th~ national sample, the mean blood lead level used here is probably low. Since the relationship between blood lead and CVD is log linear, an equivalent reduction in blood lead will have a larger impact en 33

CVD when using a lower mean blood lead level. Therefore, an

upward bias has been introduced. Sensitivity analysis, however, indicates that this bias is relatively small.

A second bias arose from th• way we calculated blood

pressure changes. When using the regression equations to

estimate blood pressur~, we sustituted the mean values for all

the variables. In ord•r for such a procedure to produce unbiased

reuslts, there would have to b• no correlation between th•

independent variables. Since correlations probably ~o exist, the

results are biased. Without knowing the correlations, though, it

is impossible to determine the direction of bias.

In the absence of good health data specifically for Boston,

simplifications are necessary in order to perform th• analysis.

The problem of bias, however, is relaiively small and

overwhelmingly outweighed by both the health effects and

population groups excluded from the analysis. Nevertheless, we · will perform sensitivity analysis irr a later section to

compensate for various weaknesses and test the strength of the results.

MATERIALS BENErITS

In addition to lower lead concentrations in drinking water

and its associated health benefits, additional treatment of

corrosive water will further reduce materials damage. This benefits category is important when we consider the enormous

investment in capital plant associated with water $Upply. Aside 34 from the petrochemical and electric power supply industries, the w~ter supply and wastewater treatment industry represents the larg•st value in capital plant. Corrosion damage is particularly acute in consumer systems where water velocity has high variability, pipes are smaller, and temperatures are higher, all characteristics which increase corrosion rates. C,:,nsumer systems also experience galvanic corrosion associated with the combination of lead sofder and copper ~ipes. Be-,: a Lt se- ,:, f th tt- large capital invsetment, enormous savings can be achieved from extending the life of pipes and other components such as hot water heaters and air conditioning systems.

8ecaus~ corrosion damage has not been examined in Boston, this analysis relies heavily on a study of Seattle performed by

Kennedy Engineers in the late 1970s. Using this study is appropriate because Seattle and Boston have very corrosive water supplies. We ~re also hampered by the lack of direct measures of corrosion rates for Boston. Consequently, the analysis uses lead concentrations as a proxy.

In 1978, Kennedy Engineers estimated annual corrosion costs to be $7,400,000: $7,Q00,000 for consumers and $400,000 for the water utility or $22.68/capita in 1985 dollars.~•

Furthermore, the National Bureau of Standards btt-lieved that corrosion control techniques used in the 1970s could reduce the costs of materials damage by 20%. Based ,:,n t~\i s .. jLtd

56 R~der, Journal of the American Water Works Association, May 1 •;ieo, p • 283. 35

To calculate the additional damage costs avoided by using

the treatment discussed in this paper, it is necessary to

estimate the incremental reduc~ion in corrosion. Using the Region

I lead concentration data as a proxy, I estimate that an

additional 25% reduction in corrosion will be achieved.e7

Applying thi~ incremental improvement to the savings ~!ready

a,: hi eved, we can expect an additional Sl.13/capita reduction in

materials damage annually from the new treatment or $635,000 for

UH/' ,:ity.

Benefits can also be derived in a second way. This weth,:,d

utilizes the conclusions reached by the AWWA concerning ~orrosion

control and avoidable costs. According to the AWWA, the

effectiveness of corrosion control ranges from 30% to 90%, while

t~,e ,:,:,rrc,si,:,r, ,:,:,sts may be Y-edu,:ed by 15% ti:, 50%. l!!.8 Usir1g the

Region ! data and assuming the prescribed treatment will maximize

corY-osion control, the ~ffectiveness of control will increase

from 75% to 90%. Assuming a lineal'" relationship between corY-osion control and reduced materials damage, this 15% improvement in control will result in a 9.75% reduction in materials damage or $2.21/capita annually (.0975 * $22.68).

Total cost reduction in Boston would therefore be Sl,242,000. By averaging the two estimates we arrive at a benefit figure of

$939,000 annually for reduced materials damage.

57 f-,::aral*kas, "C,:,ntr,:,l ,:,f Lead, Cc,pper, arid Ir,:,n PipE- C,:,rr,:,si,:,n in Boston'', p. 93. NaOH treatment reduced the mean standing grab l*ad concentration by .087 mg/L. Additional treatment will probably· reduce the concentration by anoth&r .022 mg/L which represents an additional 25% improvement.

58 AWWA, op.cit., p. 590. 36

RESULTS AND DISCUSSION

To complete the analysis, we aggregated the costs and

benefits and compared them to derive the net benefit. We also

performed some sensitivity analysis to test the strength of the

results and to compensate for some of the uncertainties arising

from the data and the analysis. Fina~ly, we considered how

~epresentative these results are for the nation as a whole.

On the treatment side, based on the judgment that the MWRA

will employ additional pH adjustment and will add zinc

orthophosphate, the utility will incur total costs of

approximately $1,600,000. Because all MWRA consumers would reap

the benefits of additional treatment, it is reasonable to

apportion costs to Boston based on its consumption.~9 Accordin9

to the MWRA demand projections, Boston consumes 43% of the

region's water. 60 Therefore, Boston's share of the treatment

costs based on consumption, would be $700,000. The utilization

of less expensive methods could dramatically lower this figure.

For example, if additional pH adjustment alone was sufficient,

the costs for Boston would be only $45,000.

59 Suburban residents would certainly see a reduction in materials damage. On the health side, benefits would proba~ly be smaller. Lead materials, however, are found in other member communities so that additional treatment would lower lead ~oncentrations in these other communities.

60 M~~rop~litan District Commission, Water demand projections, Pr&par~d by Wallace, Floyd, Associates Inc., January 1983, p. 16. 37

With respect to benefits, the analysis has examined three

major categories: avoided neurological damage in children,

reduced CVD in adult white men aged 40-59 ,and reduced materials

damage. When added together, these benefits for Boston are

S7,900,000 in 1988 increasing to $8,700,000 in 1992 (see table

IV). When the aggregate costs and benefits are compared, this

analysis indicates that additional corrosion control in Boston

will produce a positive ~et benefit of S7,200,000 in 1988,

increasing to $8,000,000 in 1992. In terms of benefit/cost ratios, the result are 11.5:1 increasing to 12.7:1.

Because much of the work on the relationship between CVD and

lead has yet to be. replicated in other analyses, we should also

consider the benefits excluding that category entirely in which

case benefits are $1,300,000 in 1988, a value which increases

slightly over the five year period. Excluding the CVD be~~fits, the analysis still yields a positive annual net benefit of

$600,000 in 1988.

Since our calculations involve a number of stages (i.e.

estimating water lead reduction, linking water lead .to blood

lead, etc) and uncertainities at each stage, an exhaustive

sensitivity analysis could generate a huge number of possible

outcomes. To simplify matters we developed only two alternate cases: the pessimistic cas~ with highest costs and lowest

benefits and the optimistic case with lowest costs and highest benefits.

For our optimistic alternative, we used the AWWA treatment cost estimate. Driven by lower chemical c0sts, TABLE IV

TOTAL BENEFITS

1':/8':/ 1'3'30 1'391 1'3'32

1:h1 ldr~n $33'3, 4'32 $342,210 $345, 2'32 $:347,351 $348, 85'3 $217, 0'36 r11:i,•c1card1 al infarcti,::,ns $1 '33,410 $1'37, 728 ·$201, 682 $20'3, 074 str,::ik11:-s $114,642 $117, 1':'4 $11'3, 534 $123, '310 '.J;128, 661 de,a.ths $6,278,613 $6,418,852 $6,547,245 ·$6, 7:37 , 263 $7,047, 68':l mat~Yials damag~ $':/3'3 I 000 $'33'3 I 000 $'3391000 ·$'33'3 1(l(H) $'33'31000

total w/,::, CVD st,2781 492 s1,2s1,210 $1,284,292 $1,286,351 s1,2s7,B59

total $7,865,157 $8,014,383 $8,152,753 $8,406,599 $8,681,305 TABLE V

COST BENEFIT COMPAf.:ISION

1 '388::..... ___1::.."3::.:8=-':J=-· -----=-l '.:.3'.:.3<.;:;.) ___--:.l..:'3.:::'3..::.1 ______.1"'"'3 .... •)=2 no;..t t,:,tal $7, 17'3, 176 $ 7, 32"3, 003 $7,466,773 ·$7, 720 , E, 18 ·$ 7 1 •3•35, 324 b/,: r.ati,:, 1 ·-::• .., 11.5 11. 7 11. '3 12.3 -• I t1€-t •,J/ ,:, CVD $5"32, 511 $5"35, 22'3 $5"38, 311 ·$600, 371 $601,-878 b/,: rati,:, w/,:, i •::i cvo 1. "3 1.'3 1. '3 . ~ 1.'3 38

Boston's share of annual treatment cost would be $550,000. To calculate benefits we used the following assumptions: apply water l•ad data to 70% of the population; use the high 95¼ confidence interval for water lead reduction - .026 mg/L; and the high estimate for materials damage - $1,242,000. Under these assumptions we calculated annual benefits to be $11,500,000 for Boston in 1988, rising to $12,775,000 in 1992 (see table VI).

In the pessimistic case, we based tYeatment cost estimates on EPA's high phosphate feed rate assGmption. Boston's share of the region's annual cost would be $860,000. On the benefits side, we employed the following assumptions to develop the low estimate: apply water lead data applies to only 40% of the population; use the low 95¼ confidence interval for water lead reduction - .018 mg/L; and the low estimate for materials damage

- $635,000. This scenario yielded $3,850,000 in annual net benefits for Boston in 1988, rising to 94,330,000 in 1992 (see table VII).

Even under pessimistic assumptions, the Boston case study clearly confirms the findings of EPA's earlier studies of the whole country: in terms of health and materials benefits, corrosion control is a worthwhile investment. What is even more impressive about these results is the fact that they represent the marginal effects. Like many pollution control activities, corrosion control is often characterized by decreasing marginal ben•fits. This implies that the benefit per dollar of treatment declines as additional treatment is implemented. Therefore, in

Boston, a city that already controls corrosion somewhat, the !ABLE VI

TCJT AL BENEFITS

1·::188 1'38"3 1'3'::10 1 ':1'31 1 ':192

,:hi ldre-t1 $4'35, 375 $4'3'3,544 $5t)4, 264 1507 1 (H)2 $50'3, oo:3 my,:11:ardi al 1nfari:ti,:,ns $302,508 $30'3, 27'3 $315,470 $327,045 ·$:32·~· ., 5•3•3 ·:;tr ,:,k e-s $17"3, 23"3 $183,238 $186, '301 $1'33, 74'3 ~201,1:31 de,aths $'3, 820,761 $10,040,666 $10,241,6'::!0' $10,617,533 $ii, 025, 128 mate,r i al s damage, $1.242.000 $1.242.000 $1.242.000 $1 r 242, c)(H) $1,242.000 total $12,039,883 $12,274,726 $12,490,325 $12,887,329 $13,316, '311

COST BENEFIT COMPAF.:ISION

1'388 1'38"3 1'3'30 1':r3: l 13':'2

t1 e-t t,:otal $11 , 4'3'3, 866 $11, 734, 70'3 $11, '350,308 $12,347,3!2 ·$ 12, 775, :3•33 1 •::. .• b/,: rati,:, 17.6 17."3 18.2 18.8 ~ . ..,. TABLE VII

TOTAL BENEFITS

1988 1'38'3 l '3'30 1 "3'31 1·3•~2

,:h1ldri:n $180, 0'36 $181, 63'3 $183,401 ·:li184,343 $185,021 myocardial infarctions $114,594 $117, 146 $11"3, 487 'li123, 862 ·u:::s, s12 str,,ke; $67, '351 $6"3, 45'3 l70,846 $73,436 $76,251 d(:-aths $3, 71"3,837 $3,802,728 $3,878,725 $4,020,778 ·$4, 174,'382 mat"1!-n.als damage $635 p (ll)!) $635,000 $635.000 $635,000 $635,000

total $4,717,477 $4,805,973 $4,887,459 $5,037,419 $5,199,866

COST BENEFIT COMPAF.:ISION

1'388 1"38'3 1 '3'31 1•::.•::.··;,--- n~t total $3,848,092 $3, '336, 587 $4,018,073 $4,168,034 $4,330,480 b/c ratio 6.9 7.0 7.1 7.3 7.6 3'3

marginal benefits generated by additional treatment will be much

lower than the marginal benefit for communities that do not

,:urrently treat their water. On the other hand, health benefits are more pronounced because the presence of lead in the water

distribution system is more widespread in Boston than in many other communities.

The discussion of marginal benefits, though does raise one

problem not'adequately addressed in this paper. Whil• this analysis has examined the combined benefits of additional and

better controlled pH adjustm•nt and zinc orthophosphate, it was

not possible to isolat& the incremental benefits produced by each

We are therefore left with the small

possibility that the vast proportion of lead reduction results

from the relatively inexpensive pH adjustment, while very rittle

results from the relatively expensive orthophosphate treatment.

Consequently, we ca~ conceive of instances where large

additional investments in treatment are required to meet the

contemplated MCL but are not warranted in terms of the health and

materials benefits they yield. In such instances, EPA should consider granting a waiver. Otherwise, inflexible enforcement of the MCL may force water utilities to invest additional resources to treat corrosion related contamination when such resources could be directed at other problems to yield greater water quality benefi;s. 40

RECOMMENDATIONS

My analysis shows that EPA should lower the MCL for lead from .05 mg/1 to .01 mg/1. The measures necessary to accomplish this task are relatively inexpenslve and the technologies are simple. The potential benefits are large and extend beyond the health effects of lead exposure. Corrosion control will lower bh• level of other toxic ~nd nuisance contaminants and reduce materials damage.

Along w~th the change in the MCL, EPA should consider instituting a waiver system. While all communities that have corrosion related lead contamination of their drinking water should undertake corrosion control bo reduce lead concentrations,

EPA should be sensitive to mariginal costs and benefits, and adopt a seccndary standard of perhaps .015 mg/L. If a community achieves this standard and can demonstrate that additional improvements would not be warranted in terms of treatment costs, a waiver from the .01 mg/L standard might be indicated. The latter requirement will be increasingly difficult to meet, though, as additional health effects of lead are better understood and monetized (i.e. renal damage) Consequently, waivers would only be granted in cases where a community just b~rely e~ceeds the MCL and further reductions in lead would require significant new investm•nts in corrosion control.

Finally, EPA, through its regional offices, should provide t~chnical assistance to cities and municipalities. Before a community chooses an appropriate treatment method it must first 41

analyz• its supply wat•r and surv•y th~ typ•s and •xtent of

mat•rials found in its distribution system. Communities must also institut• a compr*hensive monitoring plan in ord•r to evaluate the effectiveness of the corrosion control and to make

EPA should assist utilities with the necessary studies and help them dev•lop treatment strategies.

*U.S.GOVERNMENT PRINTING OFFICE •1988-516-002/80150